University of Lethbridge Research Repository
OPUS http://opus.uleth.ca
Theses Arts and Science, Faculty of
2013
Strings on complex multiplication tori
and rational conformal field theory with
matrix level
Nassar, Ali
Lethbridge, Alta. : University of Lethbridge, Dept. of Physics and Astronomy
http://hdl.handle.net/10133/3578
Downloaded from University of Lethbridge Research Repository, OPUS
brought to you by COREView metadata, citation and similar papers at core.ac.uk
provided by OPUS: Open Uleth Scholarship - University of Lethbridge Research Repository
STRINGS ON COMPLEX MULTIPLICATION TORI AND
RATIONAL CONFORMAL FIELD THEORY WITH
MATRIX LEVEL
ALI NASSAR
Master of Science, Brandeis University, 2009
A Thesis
Submitted to the School of Graduate Studies
of the University of Lethbridge
in Partial Fulfilment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
Department of Physics and Astronomy
University of Lethbridge
LETHBRIDGE, ALBERTA, CANADA
c⃝ Ali Nassar, 2013
Dedication
To the soul of my father.
iii
Abstract
Conformal invariance in two dimensions is a powerful symmetry. Two-dimensional
quantum field theories which enjoy conformal invariance, i.e., conformal field theories
(CFTs) are of great interest in both physics and mathematics. CFTs describe the
dynamics of the world sheet in string theory where conformal symmetry arises as a
remnant of reparametrization invariance of the world-sheet coordinates. In statistical
mechanics, CFTs describe the critical points of second order phase transitions. On
the mathematics side, conformal symmetry gives rise to infinite dimensional chiral
algebras like the Virasoro algebra or extensions thereof. This gave rise to the study
of vertex operator algebras (VOAs) which is an interesting branch of mathematics.
Rational conformal theories are a simple class of CFTs characterized by a finite
number of representations of an underlying chiral algebra. The chiral algebra leads
to a set of Ward identities which gives a complete non-perturbative solution of the
RCFT. Identifying the chiral algebra of an RCFT is a very important step in solving
it. Particularly interesting RCFTs are the ones which arise from the compactificatin
of string theory as σ-models on a target manifold M . At generic values of the geo-
metric moduli of M , the corresponding CFT is not rational. Rationality can arise at
particular values of the modui of M . At these special values of the moduli, the chiral
algebra is extended. This interplay between the geometric picture and the algebraic
description encoded in the chiral algebra makes CFTs/RCFTs a perfect link between
physics and mathematics. It is always useful to find a geometric interpretation of a
chiral algebra in terms of a σ-model on some target manifold M . Then the next step
is to figure out the conditions on the geometric moduli of M which gives a RCFT.
In this thesis, we limit ourselves to the simplest class of string compactifications,
i.e., strings on tori. As Gukov and Vafa proved, rationality selects the complex-
multiplication tori.
On the other hand, the study of the matrix-level affine algebra Um,K is motivated
iv
by conformal field theory and the fractional quantum Hall effect. Gannon completed
the classification of Um,K modular-invariant partition functions. Here we connect the
algebra U2,K to strings on 2-tori describable by rational conformal field theories. We
point out that the rational conformal field theories describing strings on complex-
multiplication tori have characters and partition functions identical to those of the
matrix-level algebra Um,K . This connection makes obvious that the rational theories
are dense in the moduli space of strings on Tm, and may prove useful in other ways.
v
Acknowledgements
I would like to express my deep gratitude to my supervisor Mrak Walton for introducingme to the subject of rational conformal field theories and for his kind help, patience, andcontinuous support during my PhD. This work wouldn’t have been done without his guid-ance and friendly discussions. I benefited a lot from his deep insights and comprehensiveexpertise in the field of 2D CFTs. I am grateful to Saurya Das for being supportive and formany discussions during the journal club and for his graduate course on theoretical physics.
I would like to thank Arundhati Dasgupta, Ali Keramat, and Ken Vos for their graduatecourses on theoretical physics. I am also thankful to Dan Ferguson for his inspiring way ofLab teaching which made my teaching assistant job a lot of fun.
I am indebted to Sheila Matson for all her help and encouragement and for taking careof many administrative details.
Finally, I am grateful to my friend Ahmed Farag Ali for all his help especially in movingto the University of Lethbridge and for going the extra mile for me on many differentoccasions.
vi
Table of Contents
Approval/Signature Page ii
Dedication iii
Abstract iv
Acknowledgements vi
Table of Contents vii
1 Introduction 1
2 Rational conformal field theory 18
2.1 One-loop partition function . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.2 Rational conformal field theory (RCFT) . . . . . . . . . . . . . . . . . . . . 30
2.3 Lattices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.4 Toroidal compactification . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.5 Toroidal CFTs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
2.6 c = 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.7 c = 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.8 Matrix-level affine Kac-Moody algebra . . . . . . . . . . . . . . . . . . . . . 51
3 Complex multiplication 56
3.1 Elliptic curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
3.2 Quadratic number fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.3 The endomorphism ring of elliptic curves . . . . . . . . . . . . . . . . . . . 66
3.4 Complex multiplication for higher-dimensional tori and Calabi-Yau manifolds 68
3.5 Quadratic Forms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4 Strings on CM tori 73
4.1 Matrix-level affine algebras . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
vii
4.2 Strings on a CM torus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.2.1 The Gauss product . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
4.3 Wen topological order . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.4 Rational points on Grassmannians and CM tori . . . . . . . . . . . . . . . . 88
5 Conclusion 92
Bibliography 96
viii
Chapter 1
Introduction
The two pillars of modern physics are quantum field theory and general relativity. The
former describes the interactions between elementary particles at subatomic scales.
The latter describes the gravitational attraction between macroscopic objects on large
scale. These two seemingly incompatible theories describe most phenomena in our
universe from subatomic scales to cosmological scales. String theory is an attempt to
reconcile the two theories into a unified theory of everything. By replacing the notion
of a point particle by a string, a consistent quantum theory of gravity emerges which
reproduces Einstein’s equations as its first approximation. The other forces and fields
of the standard model can also arise in a similar manner.
Two-dimensional conformal field theories (CFTs) are the building blocks of per-
turbative string theory. CFTs are also important because they describe statistical
mechanical systems with second order phase transitions at criticality. They are solv-
able non-perturbatively, and so may teach us about the non-perturbative physics of
other field theories.
The main subject of this thesis is rational conformal field theory. This is a par-
ticular class of quantum field theories which enjoys a large amount of symmetry. In
1
1.0. CHAPTER 1. INTRODUCTION
this chapter, we discuss the idea of symmetry and how it is realized and described in
physical systems.
Symmetry principles played a very important role in the study of physical phe-
nomena. Most physical systems possess a certain amount of symmetry which is
manifested in the shape or the dynamics of the system. The ideas of symmetry were
paramount in the construction of general relativity, the standard model, and in string
theory. A system is symmetric if it doesn’t change after we perform a symmetry
transformation on it. For example, an n-gon will look the same if we rotate it by
2πl/n(l = 0, 1, · · · , n− 1) around an axis through its center and perpendicular to its
plane. Also, the Hamiltonian of the hydrogen atom depends only on the distance r
between the electron and the proton and is an example of a system with spherical
symmetry. A symmetry transformation can be continuous (e.g., translation, rotation,
Lorentz boost) or discrete (e.g., space inversion or parity, time reversal, and lattice
translation).
The mathematical description of symmetry and its realization in physics is done
via group theory [1]. A set of symmetry transformations S = ⟨a, b, . . .⟩ form a group
if there is a product · which satisfies
• The product of any two transformations gives another transformation, i.e., the
set is closed under the product.
• The product is associative, (a · b) · c = a · (b · c).
• There is a unit element e ∈ S which amounts to the identity transformation or
simply doing nothing to the system.
• For every element a ∈ S there exists an inverse a−1 ∈ S, i.e., a−1 · a = e.
January 23, 2014 2 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
A group is Abelian if the order of the transformations doesn’t matter, a · b = b · a. If
the order matters a · b = b · a; then we are talking about non-Abelian groups.
It is clear that translations, rotations, and Lorentz boosts are examples of con-
tinuous groups. They are subgroups of the Poincare group ISO(1, 3), which is the
symmetry group of Minkowski space-time. The symmetry transformations of the
Poincare group can be implemented by transforming the space-time coordinates of
the system in the following way:
xµ → Λµνx
ν + aµ, (1.0.1)
where Λµν and aµ are constants which encode the parameters of the transformation.
For aµ = 0, one recovers the Lorentz group SO(1, 3) which is the group of rotations
of the space-time (spatial rotations and Lorentz boosts.)
The Poincare group is the set of space-time coordinate transformations which
preserve the relativistic infinitesimal interval
ds2 = ηµνdxµdxν = η′µνdx
′µdx′ν , (1.0.2)
where ηµν is the metric of space-time. The invariance of ds2 imposes
ηµν = ΛαµΛ
βνηαβ, aµ = constants. (1.0.3)
From the above equation, we learn that the Poincare group is the set of space-
time coordinate transformations which leave the metric form invariant. It is a 10
parameter group (4 translations, 3 spatial rotations, and 3 Lorentz boosts) which
embodies the relativistic description of nature. We remark that any fundamental
physical theory must remain invariant under the Poincare group as required by the
relativistic symmetry of the fundamental laws of physics. This relativistic symmetry
January 23, 2014 3 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
signifies our ability to repeat experiments at different places, at different times, in
different directions, and in different inertial reference frames without affecting the
outcome of the experiment.
If we relax the condition of the form invariance of the metric and demand that the
metric remains form invariant up to an overall function, we arrive at the conformal
group of space-time. Invariance under the conformal group, however, is not always
guaranteed and most physical systems break conformal invariance. An example of
a conformal transformation which is not a Poincare transformation is scaling which
acts as xµ → λxµ. One would expect that scaling the dimensions of a physical system
will lead to different outcomes of physical experiments. For example, by doubling
the volume of a closed room the pressure will decrease (assuming the ideal gas law
is valid). Invariance of a physical system under scaling or more generally under the
conformal group requires intricate dynamics. One can get an idea of what kind of
physical systems might enjoy conformal symmetry by looking at (1.0.2) for the case
of light rays, i.e., the null interval ds2 = 0. In this case, scaling the metric would
leave the interval null. Hence light propagation or more generally massless particles
propagation is conformally invariant. It turned out that classical Maxwell and Yang-
Mills theories in 4D are conformally invariant.
The symmetries we have been discussing about so far are external symmetries
acting on the background space-time in which the physical system lives. There is,
however, a different kind of symmetry which corresponds to transformations of the
internal degrees of freedom of the system. These internal symmetries are not directly
related to space-time. An example of a continuous internal symmetry is the phase
January 23, 2014 4 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
rotation of the wave function in quantum mechanics
ψ(x) −→ eiαψ(x), (1.0.4)
where α is a constant phase. The set of constant phase transformations in (1.0.4)
form the global U(1) group.
Another example is the isospin invariance of nuclear interactions. Nuclear forces
between protons and neutrons does not depend on the electric charge and is blind
to whether the particle is a proton or a neutron. The proton and the neutron are
treated two different states of the same particle, the nucleon, with different isospin
charge (P
N
)−→ S
(P
N
), (1.0.5)
where S is a 2 × 2 constant unitary matrix with unit determinant. This is the
global SU(2) group which played a very important role in the early days of the
strong-interaction physics. In terms of the quark structure of hadrons, the SU(2)
symmetry of strong interactions is due to the approximate equality of the masses of
the up and down quarks. If one also include the charm quark and ignore the small
mass differences with the up and down quark, we arrive at the SU(3) flavor group.
This group underlies the eight-fold way scheme of the classification of baryons and
mesons [2]. One can also have discrete internal symmetries like charge conjugation
which replace every particle with its antiparticle.
One can allow the above symmetry transformations to depend on space-time by
having independent symmetry transformations at each point in space-time, i.e., by
gauging the global symmetries. Such local symmetries are called gauge symmetries,
and the theories which describe them are called gauge theories. Such theories are
January 23, 2014 5 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
known generically as Yang-Mills theories [3, 4] and they describe the interactions
between massless spin-1 gauge bosons. The modern description of the fundamental
interactions of nature relies on the gauging of the internal symmetries they respect.
The Maxwell theory of the electromagnetic interactions is a U(1) gauge theory. On
the other hand, the weak and the strong interactions are based on gauging the SU(2)
and the SU(3) groups, respectively. The standard model of particle physics which
describes the non-gravitational forces of nature is based on a Yang-Mills theory with
a gauge group SU(3)× SU(2)× U(1).
Einstein’s theory of gravity can be recast as the gauge theory of the Poincare
group. The classic example is the 4-potential Aµ of classical electrodynamics, where
the U(1) gauge symmetry tells us that Aµ and Aµ+∂µΛ give the same field strength,
i.e., the same physics. In a gauge theory, physical observables must be gauge invariant,
i.e., they transform trivially under the gauge group. We should stress that gauge
symmetries are not symmetries in the usual sense but rather they are redundancies
in the description of physical systems. In the case of gravity, where the gauge degrees
of freedom are the space-time coordinates, this implies that local operators are not
gauge invariant. In the standard model, gauge symmetry is realized as an internal
symmetry acting on fields which live in a fixed space-time background. In gravity,
gauge symmetry is an external symmetry acting on the space-time metric which is
a dynamical field. This distinction between the way gauge symmetry is realized in
the standard model and in gravity is one of the reasons why gravity is still hard to
reconcile with the principles of quantum mechanics.
Supersymmetry (SUSY) is an extension of space-time symmetries which relates
bosons and fermions [5, 6]. It predicts the existence of a superpartner with the same
January 23, 2014 6 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
mass for every elementary particle in nature. Since no superpartners have been dis-
covered so far, it is assumed that SUSY is broken at low energies which uplifts the
degeneracy between bosons and fermions masses. One of the motivations for SUSY
is that it gives a resolution of the hierarchy problem of the Standard Model. Without
the extra superpartners, quantum corrections drives the Higgs boson mass close to
the Planck mass and a very superficial fine tuning is required to keep the Higgs mass
low. The mathematical description of SUSY requires the introduction of fermionic
coordinates θα. These fermionic coordinates transform as spinors under the Lorentz
group and together with the bosonic space-time coordinates xµ they make up the
superspace. SUSY transformations can be realized as translations and rotations of
the superspace [7]. The relevant CFTs which arise in string theory are the ones which
enjoys a certain amount of supersymmetry. These are superconformal CFTs which
underlies the superstring perturbation theory.
To appreciate the importance of symmetry in the description of physical systems,
we recall the Noether theorem: For every global continuous symmetry, i.e., a trans-
formation of a physical system which acts the same way everywhere and at all time,
there exists an associated time-independent quantity (i.e., a conserved quantity). Ap-
plying the Noether theorem to space-time translations leads to the energy-momentum
conservation. Similarly, invariance under spatial rotations leads via Noether theorem
to the conservation of angular momentum. Lorentz boosts, however, don’t lead to
any conserved quantity since they don’t commute with time evolution. The Noether
theorem can also be applied to internal symmetries. The global U(1) symmetry of the
electromagnetic interactions leads to the conservation of electric charge. The global
January 23, 2014 7 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
SU(2) symmetry of the weak interaction and the global SU(3) of the strong interac-
tions leads to the conservation of global charges as well. Gauging a global symmetry
doesn’t lead to any new conserved charges. This supports the claim that local gauge
invariance is not a symmetry in the usual sense. SUSY leads to conserved fermionic
supercharges as well. These supercharges transform bosons into fermions and vice
versa.
Classically, it is straightforward to tell whether a system is invariant under a
certain symmetry by looking at the action
S =
∫L[qi(t), qi(t)]dt, (1.0.6)
where L[qi(t), qi(t)] is the Lagrangian of the system. A symmetry of a classical system
is a transformation of the dynamical variables qi(t)→R[qi(t)] that leaves the action
unchanged. In the Hamiltonian formalism, a symmetry is realized by a generator
which has a vanishing Poisson bracket with the Hamiltonian
IR, H = 0. (1.0.7)
This formalism can be generalized to classical field theories which are described by a
Lagrangian density L(ϕi(x), ∂µϕi(x)).
In the quantum theory we no longer have the classical configuration space but
rather a Hilbert space of quantum states. The classical symmetry is realized linearly
by unitary operators acting on the Hilbert space. It could happen that a classical
symmetry doesn’t survive after quantization. In this case, we speak of an anomalous
symmetry. To see how this could happen, let us recall that the quantum theory is
described by an amplitude which can be given as a path integral
Z =
∫DqeiS[q,q], (1.0.8)
January 23, 2014 8 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
where S[q, q] is the classical action written in terms of the c-number classical variables.
The only way a classical symmetry could be broken is if the measure Dq fails to be
symmetric. This can happen if the regularization prescription, which is needed to
make sense of the above path integral, violates the symmetry. An example of a
theory which is conformally invariant (classically) is Yang-Mills theory in 4D (e.g.,
the Maxwell theory of electromagnetism). However, the conformal symmetry in this
case doesn’t survive after quantization and the conformal symmetry is anomalous.
Yang-Mills theories in 4 dimensions are conformally invariant but quantum Yang-
Mills theories are not. It is known that classical Yang-Mills theory which describes
massless particles develops a mass gap due to quantum mechanical corrections [8].
An example of a theory which is conformally invariant even on the quantum level is
N = 4 super Yang-Mills theory in 4D.
Another important feature about the way symmetry is realized in quantum field
theory is the concept of phases. Quantum field theory can have different phases
in which the symmetry of the fundamental Lagrangian is realized differently. For
example, the high energy SU(2) × U(1) symmetry of the electroweak interactions
is spontaneously broken in the Higgs phase at low energy to U(1) symmetry. This
however doesn’t mean the symmetry is lost. It just means that the ground state of
the system doesn’t respect the symmetry.
The way symmetry is realized in string theory is different from point-particle
quantum field theories. The fundamental objects in perturbative string theory are
one-dimensional strings which can be open or closed. When the string moves in space-
time it sweeps out a two-dimensional surface. This surface is called the world-sheet
of the string.
January 23, 2014 9 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
Figure 1.1: The classical trajectory of a string minimizes the area of the world sheet.
The surface is specified by the functions
Xµ = Xµ(σ0, σ1), (1.0.9)
which describe the embedding of the string in the background space-time.
The starting point of perturbative string theory is the world-sheet action of a
string moving in some background space-time with a metric g. The embedding of the
string world-sheet Σ in a background D-dimensional space-time M is described by
the maps X : Σ→M . The action of the bosonic string is
S(X, g) =1
4πα′
∫Σ
d2x√hhαβgµν(X)∂αX
µ∂βXν , (1.0.10)
where hαβ is the world-sheet metric and α′ is the only dimensional parameter of string
theory and has the dimensions of length squared.
The above action can be supersymmetrized which leads to an N = 1 supergravity
theory on the world sheet. After gauge fixing it will give a theory with N = 1
superconformal symmetry. We will only consider bosonic CFTs in this thesis. We
now specialize to flat space-time backgrounds gµν = δµν with Euclidean signature.
January 23, 2014 10 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
Diffeomorphism plus Weyl invariance of the above action enables one to chose [9]
hαβ = δαβ. (1.0.11)
This is always possible locally on any world-sheet. The above action now boils down
S(X) =1
4πα′
∫Σ
d2x ∂αXµ∂αX
µ, (1.0.12)
This is the action of a two-dimensional quantum field theory living on the world-
sheet where the string coordinates Xµ act as D massless bosons in two dimensions.
The above action is scale invariant. If we scale τ → λτ and σ → λσ the action
remains the same. The action is invariant under the much bigger conformal group
in two dimensions and the two-dimensional quantum field theory which governs the
fluctuations of the world sheet is conformal. The space-time physics of string theory,
in particular the propagation of strings and their interactions, can be pulled back to
the two-dimensional conformal field theory (CFT) on the world-sheet. The geometric
characteristics of the target manifold follow from the properties of the world-sheet
CFT. This world-sheet CFT is required to have a vanishing conformal anomaly1.
As it turns out, the vanishing of the conformal anomaly translates to the Einstein’s
equations in 25 D in the case of the bosonic string and 10 D for the supersymmetric
string. Thus a choice of a CFT on the world-sheet is equivalent to a choice of a
background (or a classical solution) in string theory. The classification of 2D CFTs
is equivalent to classifying the classical solutions of perturbative string theory.
There are five supersymmetric string theories in ten-dimensional flat space [9,10]:
Type I with an SO(32) gauge group, type IIA and type IIB with no gauge symmetry
and Heterotic SO(32) and Heterotic E8×E8 with gauge groups SO(32) and E8×E8,
1The vanishing of the conformal anomaly is required in order to preserve the 2D Weyl invariance.
January 23, 2014 11 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
respectively. One can engineer more general gauge groups by the inclusion of D-branes
(for a comprehensive review see [11]). D-branes are hyperplanes on which open strings
can end. They are dynamical objects in string theory which arise upon quantizing
open strings subject to Dirichlet boundary conditions. In the world-sheet description
of string theory, D-branes give rise to boundary conformal field theories [12–14].
Gauge symmetry can also arise by compactification on manifolds with isometries [15]
or from singularities in the compactification manifold [16, 17]. The aforementioned
gauge groups have an elegant world-sheet description in terms of spin-1 currents which
generate an affine Kac-Moody algebra [18–21]. In this thesis, we will study similar
algebras which arises from compactification of strings on tori.
It is worth mentioning that there are no global symmetries in string theory. It can
be shown using a simple world-sheet argument [9, 10] that any symmetry in string
theory must be gauged. This is consistent with the observation that there can’t be
global conserved charges in a theory with black holes. Global charges will disappear
behind the horizon and there is no way for them to be measured. However, if the
global symmetry is gauged, then one can measure the charge while staying away
from the horizon by measuring the flux of the field strength. This doesn’t violate
the No-hair theorem since the charges which characterize a black hole correspond to
a global symmetry which is gauged [22]. The fact that string theory doesn’t allow
global symmetries makes it a viable candidate for a theory of quantum gravity.
In studying the simplest compactifications of string theory one considers a space-
time which has product form
M10 =M3,1 ×X6, (1.0.13)
where M3,1 is our Minkowski space-time and M6 is the a compact six-dimensional
January 23, 2014 12 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
space.
In terms of the CFT description of the string
CFT10 = CFText4 × CFTint
6 , (1.0.14)
where CFTint6 is the internal CFT which corresponds to the compactified dimen-
sions. Different choices for CFTint6 give different space-time physics. Supersymmetry
in space-times requires that the world-sheet CFT to have N = 2 supersymmetry.
Classical solutions of string theory are given in terms of N = 2 superconformal field
theories. We will not consider superconformal theories in this thesis and will only
consider the bosonic sectors. Adding the fermions will not change any of the results
in this thesis and the bosonic theories we will consider can be uplifted to N = 2
superconformal field theories which are relevant in string theory.
We will study the CFTs which result from string compactification on a two torus
T 2. Toridal compactification is the harmonic oscillator of string compactification due
to its simplicity and exact solvablity [23, 24]. A torus is a manifold with an U(1)2
isometry which corresponds to the compact translations around the two cycles of
the torus. We will describe toroidal compactification in more details in Chapter 2.
The U(1)2 symmetry gives rise to two conserved currents on the world-sheet CFT
of the string and one expects the existence of two conserved charges. These are
the quantized momenta which correspond to translations in the compact dimensions.
However, there are two other conserved charges which have stringy origins. Strings
can wind around the compact dimensions and the winding numbers behaves like a
conserved charge. These topological charges come from the dual U(1)2 currents and
they characterize the winding states. The momentum and winding charges live on a
self-dual integer Narain lattice Γ2,2 which for the case of T 2 has a signature (2, 2) [23].
January 23, 2014 13 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
We will study these lattices in detail later.
It is always useful to find a geometric interpretation of a given CFT in terms
of a σ-model on some target manifold M . The interplay between the algebraic and
geometrical pictures have been very fruitful in revealing many aspects of CFTs and
the target manifolds. This is most obvious in the case of strings on Calabi-Yau
manifolds M which are described by an N = 2 superconformal field theory. The
N = 2 superconformal algebra admits an automorphism which reverses the sign of a
U(1) charge. This operation which gives an isomorphic superconformal field theory
gives a completely different Calabi-Yau target W . The cohomology groups H1,1 and
H2,1 of M and W are interchanged under the aforementioned automorphism. This
lead the authors of [25,26] to propose the existence of a mirror manifold W for each
Calabi-Yau manifold M with H1,1(M) = H2,1(W ) and H2,1(M) = H1,1(W ). The
main motivation for this proposal is the fact that both M and W give isomorphic
superconformal field theories. An explicit construction of mirror pairs of Calabi-Yau
manifolds was given in [27].
A conformal field theory is rational when the chiral algebra generated by the
conserved currents is large enough so that the Hilbert space contains a finite number
of representations. For CFTs based on a target space M rationality can arise at
special values of the geometric moduli of M . To see how a CFT can become rational
at certain points of its moduli space, we recall the simplest example of an RCFT: the
c = 1 CFT of a compact free boson on a circle of rational square radius R2 = r/s.
The radius R plays the role of a geometric modulus. At a generic value of R, the
number of primary fields is infinite. At R2 = r/s, the chiral algebra of this theory
is extended. The infinite list of representations reconstitute themselves into finitely
January 23, 2014 14 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
many irreducible representations of the extended chiral algebra (see Chapter 2).
More interesting RCFTs arise is in string compactifications on Calabi-Yau mani-
folds at special values of their moduli, e.g., in Gepner models [28]. For example, the
Fermat quintic in CP4
P (z) =5∑
a=1
(za)5 = 0. (1.0.15)
This quintic corresponds to a Gepner point where the corresponding CFT is rational
and is described by a product of N = 2 minimal models [28].
It is important to understand the conditions for rationality for CFTs based on a
target manifold M . For example, the simplest compactifications of a string theory
are on tori Tm–when are they described by RCFTs? This question has been studied
in [29–32]. In particular, Wendland [30] derived rationality conditions valid for all
torus dimensions m. More importantly for us, in the case of 2-tori, Gukov and
Vafa [32] found a simple, geometric criterion for rationality. For T 2, the modular
parameter τ and the Kahler parameter ρ must take special values; they must belong
to an imaginary quadratic number field. Such 2-tori have the property of complex
multiplication, and are known as complex-multiplication (CM) tori. For them, a
Gauss product exists, and was used in [33] to classify the corresponding RCFTs.
We will study strings on (CM) tori which are a particular class of tori with more
symmetry. CM tori are characterized by the values of their modular parameter τ
which satisfies
aτ 2 + bτ + c = 0 −→ τ =−b+
√b2 − 4ac
2a(1.0.16)
that is, τ belongs to a quadratic number field Q(D), where D = b2 − 4ac < 0. The
quadratic number field Q(D) is the set of complex numbers z which can be written
January 23, 2014 15 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
as
z = α + β√D, (1.0.17)
where α, β ∈ Q are rational number. It is obvious that the modular parameter of a
CM torus belongs to Q(D). This is also true for the the Kahler parameter ρ which
is exchanged with τ under mirror symmetry. As was shown in [32] rationality picks
CM tori.
In earlier work, Gannon [34] gave a complete classification of the modular invariant
partition functions of the Abelian matrix-level affine algebra Um,K . This algebra is
important in the description of the Fractional Quantum Hall Effect (FQHE). The
algebra Um,K also appears in the 2D CFTs which are the holographic dual of 3D
Chern-Simons theories which describe topological membranes [35–37].
We give a geometric connection of the Um,K algebra for any K and for m = 2.
Using the Gauss product [33], we relate the U2,K algebra to RCFTs which arise from
strings on complex multiplication tori. We will uncover a connection between the
matrix-level algebras U2,K and the CM tori, and so to the extended Moore-Seiberg
algebras that make strings on them describable by RCFTs. We will use the Gauss
product [33] to construct the matrix levels for the algebra U2,K using the geometric
data encoded in τ and ρ which correspond to strings on CM tori. We will show that
the RCFTs which arise from strings on CM-tori and the RCFTs based on the Um,K
algebra have the same set of characters and the same partition function. One can
write down a Moore-Seiberg algebra for the RCFTs which result from strings on CM
tori. But as we will see in the next chapter, this algebra is already complicated for
case of a rational circle. In this thesis, we propose the algebra Um,K as an alternative
description of the extended chiral algebra for strings on CM tori.
January 23, 2014 16 Ali Nassar
1.0. CHAPTER 1. INTRODUCTION
Using this connection between the U2,K algebra and strings on complex multipli-
cation tori, we will show that RCFTs are dense inside Narain moduli space. This
confirms the results in [32] about the density of the RCFTs which arise from strings
on CM tori.
Since the algebra Um,K also appears in the study of the FQHE, we use the connec-
tion between U2,K and strings on CM tori to relate the topological order of fractional
quantum hall states to the number of D0 branes allowed on CM tori. We will also
propose a generalization of the matrix-level algebra to the non-Abelian case.
This thesis is organized as follows: In Chapter 2, we first give a brief introduc-
tion to CFT then we discuss the algebra Um,K and its modular invariant partition
functions. In Chapter 3, we give a physicist’s introduction to complex multiplication
tori and we describe various examples. In Chapter 4, we present our main results.
Chapter 5 is the conclusion.
January 23, 2014 17 Ali Nassar
Chapter 2
Rational conformal field theory
Two-dimensional conformal field theories (CFTs) [38] (for reviews see [39, 40]) are
quite well understood. They are probably the best understood among all quantum
field theories. Two-dimensional conformal systems are very special, because only in
two dimensions is the conformal algebra infinite-dimensional1. The local conformal
symmetry described by the Virasoro algebra, or one of its extensions, is in many cases
powerful enough to lead to a complete determination of the operator spectrum, as
well as to explicit formulas for the correlation functions. This constitutes a complete
non-perturbative solution of the theory.
On the physics side CFTs appear in two contexts. First, they describe the be-
havior of many statistical mechanical systems at a renormalization group fixed point.
When we look at these systems at larger and larger scales (or by going to the deep
infrared) we find that fluctuations occur at all length scales. At the critical point
(the phase transition point), the system is covariant under scale transformations. It
was shown [38,41] that scale invariance plus rotational invariance (or Lorentz invari-
ance in Minkowski signature) implies the invariance under the full conformal group.
1In higher dimensions, the conformal group is finite dimensional.
18
2.0. CHAPTER 2. RATIONAL CONFORMAL FIELD THEORY
Hence, these systems can be described by a CFT. These theories thus give concrete
instances of non-trivial fixed points of the renormalization group, many of which have
a realization in statistical mechanical systems.
An example of a critical phenomenon is the ferromagnetic phase transition at
the Curie temperature Tc in ferromagnetic materials (like iron). Above the Curie
temperature, the dipole moments of the individual atoms are randomly aligned, re-
sulting in a vanishing total dipole moment and the system is in the paramagnetic
phase. When the temperature is decreased below the Curie temperature, the individ-
ual dipole moments pick a preferred direction in order to minimize their total energy.
This results in a non-vanishing total dipole moment (magnetization) and the system
is in the ferromagnetic phase. The paramagnetic phase is a disordered phase where
the system is invariant under three-dimensional rotations, i.e., the system enjoys an
SO(3) symmetry. The ferromagnetic phases is an ordered phase and the rotational
symmetry is broken. Ordered-disordered phases are a characteristic of second order
phase transitions and it is always characterized by an order parameter which takes a
non-zero value in the broken phase. In the case of the ferromagnetic phase transition,
the total magnetization is an order parameter.
The ferromagnetic phase transition can be described theoretically using the Ising
model [42]. The 2D Ising model is a square lattice with a discrete variable si (spin) at
each lattice site. The variable si can can take only the values +1 or −1. Neighboring
spins interact with each other through an exchange coupling
E = −J∑<i,j>
sisj, (2.0.1)
where J is the exchange energy. In the high temperature phases, ⟨s⟩ = 0 and in the
low temperature phase ⟨s⟩ = 0.
January 23, 2014 19 Ali Nassar
2.0. CHAPTER 2. RATIONAL CONFORMAL FIELD THEORY
An important quantity is the correlation length ξ(T )
⟨sisj⟩ − ⟨si⟩⟨sj⟩ ∼ e−|i−j|ξ(T ) . (2.0.2)
For generic T , the correlation length ξ(T ) is finite and the statistical fluctuations of
the spins are correlated over a finite length. As T approaches its critical value, the
correlation length diverges
ξ(T ) ∼ 1
|T − Tc|and the system starts to fluctuate over all length scales (see Figure 2.1 ). At the
critical point, the correlators follow a power law in contrast to the exponential decay
in (??). This self-similar behavior and the divergence of the correlation length is a
manifestation of the scale invariance of the system at the critical temperature. It
turns out that scale invariance is a part of a bigger symmetry, conformal symmetry,
which emerges at the critical point. For statistical models in two dimensions, the
continuum description at a second-order phase transition is given by a conformal
field theory. The prime example is the Ising model which is a two-dimensional model
of ferromagnetism. The 2D Ising model was analytically solved by Onsager [43] where
he showed that correlation functions and the free energy of the model are given by a
theory of free lattice fermions.
This is an example of a second-order phase transition which separates a high tem-
perature disordered phase (with the expectation value ⟨s⟩ = 0) and a low temperature
ordered phase (with ⟨s⟩ = 0).
The second major application of CFT in physics is string theory. The classical
solutions of perturbative string theory (or the vacuum configurations) are described by
2D CFTs. Classifying and understanding these solutions amounts to understanding
the classifications of CFTs.
January 23, 2014 20 Ali Nassar
2.0. CHAPTER 2. RATIONAL CONFORMAL FIELD THEORY
Figure 2.1: A snapshot of the Ising model at Tc. There are fluctuations on all lengthscales (self-similarity) [44].
This chapter is a review of conformal field theory. We start with the definition of
conformal transformations, then we go on and develop the basic tools used in CFT,
focusing on those aspects of CFT which will be used in the thesis. Our discussion
will be based on [9, 39,40]
The conformal group is the set of transformations which preserve the angle be-
tween vectors [40]
cos θXY =X · Y√
∥X∥2√∥Y ∥2
=ηµνX
µY ν√ηµνXµXν
√ηµνY µY ν
. (2.0.3)
The generators of the conformal group are
Pµ = −i∂µ translation
D = −ixµ∂µ dilation
Lµ,ν = i(xµ∂ν − xν∂µ
)rotation
Kµ = −i(2xµx
ν∂ν − x2∂µ)
special conformal transformation.
(2.0.4)
January 23, 2014 21 Ali Nassar
2.0. CHAPTER 2. RATIONAL CONFORMAL FIELD THEORY
Now we define the generators Ja,b = −Jb,a, a, b ∈ −1, 0, 1, . . . , d as follows
Lµ,ν = Jµ,ν , J−1,µ =1
2
(Pµ −Kµ
), µ, ν = 1, · · · , d,
J−1,0 = D, J0,µ =1
2
(Pµ +Kµ
).
(2.0.5)
They satisfy the algebra
[Jab, Jcd
]= i(ηadJbc + ηbcJad − ηacJbd − ηbdJac
), (2.0.6)
where ηab = diag(−1, 1, 1, . . . , 1). This is the algebra of SO(d + 1, 1), the conformal
group in d-dimensions. The dimension of SO(d + 1, 1) is (d + 2)(d + 1)/2, i.e., the
conformal group in d dimensions is (d + 2)(d + 1)/2-parameter group. Quantum
field theories in d > 2 with conformal symmetry are interesting, and although the
conformal symmetry leads to some simplifications, it is still not big enough to give
exact solutions.
The situation in d = 2 changes dramatically, since the conditions on ϵ give
∂1ϵ1 = ∂2ϵ2, ∂1ϵ2 = −∂2ϵ1. (2.0.7)
If we use complex coordinates
z = x1 + ix2, z = x1 − ix2,
∂ =1
2
(∂1 − i∂2
), ∂ =
1
2
(∂1 + i∂2
),
ϵ(z) = ϵ1 + iϵ2, ϵ(z) = ϵ1 − iϵ2,
(2.0.8)
then (2.0.7) are the Cauchy-Riemann equations for the real and imaginary parts of
ϵ(z). Hence, conformal transformations in d = 2 coincide with the set of holomorphic
(anti-holomorphic) coordinate transformations (conformal mappings) of the complex
plane
z −→ f(z), z −→ f(z). (2.0.9)
January 23, 2014 22 Ali Nassar
2.0. CHAPTER 2. RATIONAL CONFORMAL FIELD THEORY
It is known that the set of holomorphic coordinate transformations in d = 2 is infinite
dimensional and so that the conformal algebra in d = 2 is infinite dimensional. It
is worth mentioning that the conformal group of C is not infinite-dimensional since
only the transformations corresponding to L0, L1, L−1 are globally defined (see [45]).
If, however, we work locally (or on the level of the algebra) then we do get an infinite
dimension algebra.
Locally, a holomorphic transformation can be written as
f(z) = z+ ϵ(z) = z+∑n∈Z
(− ϵnzn
), f(z) = z+ ϵ(z) = z+
∑n∈Z
(− ϵnzn
). (2.0.10)
The generators of the conformal transformations are
ln := −zn+1∂, ln := −zn+1∂. (2.0.11)
They act on the space of functions and satisfy the classical Witt algebra
[lm, ln
]= (m− n)lm+n,
[lm, ln
]= (m− n)lm+n, [ln, lm] = 0 (2.0.12)
which is a direct sum A⊕ A of two commuting subalgebras. In what follows we only
treat the holomorphic part, with the understanding that the analysis also applies to
the anti-holomorphic part. The classical Witt algebra is realized quantum mechan-
ically with a central extension (Virasoso algebra). The central term is related to a
quantum mechanical anomaly in the transformation of the energy-momentum tensor.
The Virasoro algebra Vir⊕ Vir is
[Ln, Lm] = (m− n)Ln+m +c
12(m3 −m)δm+n,0, m, n ∈ Z,
[Ln, Lm] = (m− n)Ln+m +c
12(m3 −m)δm+n,0,
[Ln, Lm] = 0.
(2.0.13)
January 23, 2014 23 Ali Nassar
2.0. CHAPTER 2. RATIONAL CONFORMAL FIELD THEORY
This is the chiral algebra which underlies any conformal field theory. It is the quantum
extension of the classical Witt algebra and is characterized by the central terms
proportional to c and c. The central terms vanish when restricted to the generators
of the global conformal group, L0, L1, L−1 and L0, L1, L−1 implying that this group is
non-anomalous. It is worth mentioning that a CFT can have a bigger chiral algebra
than Vir⊗ Vir, e.g., W -algebras, current algebras, superconformal algebras.
The relevant representations of the Virasoro algebra which are important in phys-
ical applications involve operators which have a fixed scaling dimension ∆. These are
the conformal primary operators Oi, which are eigenfunctions of the scaling operator
D with the eigenvalue ∆i = hi + hi. Here, hi denotes the holomorphic conformal
dimension and hi denotes the anti-holomorphic conformal dimension. The normal-
ization of the primary operators is arbitrary and we chose the basis of the primary
operators such that the two-point function takes the form
⟨Oi(z1, z1)Oj(z2, z2)⟩ =δij
|z1 − z2|2∆i. (2.0.14)
CFTs admit an operator product expansion (OPE) in which two local operators
inserted at nearby points can be closely approximated by a string of operators at one
of these points
Oi(z, z)Oj(z, z) =∑k
Ckij(z − w, z − w)Ok(w, w), (2.0.15)
where Oi(z) is a set of local fields which can be decomposed into a direct sum of
conformal families or conformal towers
Oi(z) =∑n
[ϕn]. (2.0.16)
On top of each one of these towers sits a highest weight state which corresponds to a
primary field. The other members of the conformal family [ϕn] are called descendant
January 23, 2014 24 Ali Nassar
2.0. CHAPTER 2. RATIONAL CONFORMAL FIELD THEORY
fields and they have conformal dimensions
h(k)n = hn + k, h(k)n = hn + k. (2.0.17)
where k, k ∈ Z by construction.
The OPE can be written down for any set of local operators in any quantum field
theory, e.g., QCD. The novel thing about CFT is the absence of any dimensional
parameter l. It was argued that the above OPE converges in a CFT since harmful
terms of form e−l/|z−w| are absent.
The OPE of the energy-momentum tensor is of utmost importance
T (z)T (w) ∼ c/2
(z − w)4+
2T (z)
(z − w)2+
∂T (z)
(z − w)+ reg
T (z)T (z) ∼ c/2
(z − w)4+
2T (z)
(z − w)2+
∂T (z)
(z − w)+ reg
T (z)T (z) ∼ reg,
(2.0.18)
where c, c ∈ N are the holomorphic and anti-holomorphic central charges, respectively.
The regularity of the T (z)T (z) OPE is a manifestation of the decoupling of the
holomorphic and anti-holomorphic sectors. They encode the quantum anomally in
the transformation of T (z) and T (z). One-loop modular invariance imposes
c− c = 0 mod 24. (2.0.19)
This condition can be realized by taking c = c as in bosonic and type II string theory.
January 23, 2014 25 Ali Nassar
2.1. One-loop partition function
The Virasoro algebra (2.0.13) is derived from the above OPE of T with itself
[Ln, Lm] =1
(2πi)2
∮0
dw
∮w
dzwm+1zn+1T (z)T (w)
=1
(2πi)2
∮0
dw
∮w
dzwm+1zn+1
(c/2
(z − w)4+
2T (z)
(z − w)2+
∂T (z)
(z − w)
)[Ln, Lm] =
1
(2πi)2
∮0
dw
∮w
dzwm+1zn+1T (z)T (w)
=1
(2πi)2
∮0
dw
∮w
dzwm+1zn+1
(c/2
(z − w)4+
2T (z)
(z − w)2+
∂T (z)
(z − w)
)[Ln, Lm] =
1
(2πi)2
∮0
dw
∮w
dzzn+1wm+1T (z)T (w).
(2.0.20)
2.1 One-loop partition function
So far, we have been studying the CFT on the complex plane, which after adding
the point at infinity, is equivalent to a sphere (the Riemann sphere, a genus h =
0 surface). The two chiral algebras (the holomorphic and anti-holomorphic) were
considered independent and in principle one can have different chiral algebras in the
two sectors. However, in order to have single-valued and monodromy-free correlation
functions on a genus h = 0 surface, one has to couple the two sectors together
and impose various consistency conditions on the irreducible representations in each
sector. It turns out that a set of strong constraints appear when one studies the CFT
on a Riemann surface with higher genus. In particular, on a Riemann surface of genus
h = 1 or a torus. The requirement of modular invariance puts a stringent constraint
on the spectrum coming from the interaction of the holomorphic and antiholomorphic
sectors [46].
The one-loop partition function, or the zero-point function on the torus, encodes
the spectrum and the multiplicity of the different primary fields which appear in the
January 23, 2014 26 Ali Nassar
2.1. One-loop partition function
CFT. In the context of string theory, the one-loop partition function is the one-loop
stringy Feynman diagram and it encodes the different fields which propagate in the
closed string channel together with their multiplicities. For application in critical
phenomena, the torus geometry appears when one considers the plane with periodic
boundary conditions in two directions. The partition function is defined as
Z(q, q) = TrH
(qL0−c/24qL0−c/24
), (2.1.1)
where
q = exp 2πit, q = exp−2πit. (2.1.2)
The parameter t is the modular parameter of the torus2.
The Hilbert space of the CFT is decomposed as
H =∑λ,µ∈E
MλµHλ ⊗Hµ, (2.1.3)
where we assumed that the CFT is compact so the spectrum of conformal dimensions
is discrete. The set E could in principle be infinite, however, when the CFT is rational
E is finite.
The one-loop partition function Z = Tre−βH can be written as
Z(t, t) =∑λ,µ∈E
Mλµ χλ(t) χµ(t), (2.1.4)
where χλ(t) is the genus-1 character of an irreducible representation hλ
χλ(τ) = TrHλ
(qL0− c
24
), (2.1.5)
where Hλ denotes the Verma module built from the highest weight state |hλ⟩.2We point out that we are using t instead of the more familiar τ since we reserve τ to denote the
modular parameter of the target space torus to be discussed later.
January 23, 2014 27 Ali Nassar
2.1. One-loop partition function
ω10
ω2 ω1 + ω2
Figure 2.2: The torus as a parallelogram with opposite sides identified.
The matrix Mλµ in (2.1.4) incorporates the multiplicities of the different fields
which appear in Z(t, t) and as such Mλµ ∈ Z+. Uniqueness of the vacuum forces
M00 = 1. Modular invariance will put more constraints onMλµ and to see that we
will need to study the geometry of tori in more detail.
The torus can be described as a parallelogram in C generated by two linearly
independent lattice vectors ω1, ω2 with the opposite sides of the parallelogram being
identified
z ∼ z +mω1, z ∼ z + nω2, m, n ∈ Z. (2.1.6)
Since we have a scaling symmetry at our disposal we can rescale one of the lattice
vectors, say ω1 to 1. The two lattice vectors will be denoted by 1, t ∈ C, where
t = ω1/ω2 is the scale invariant parameter which is relevant in the CFT. The linear
independence of ω1 and ω2 requires ℑ(t) > 0, i.e., t ∈ H+. The parameter t is called
the modular parameter of the torus and it defines the shape and the size of the torus.
Any integer linear combination of the lattice vectors will give an equivalent basis
for the lattice. One should make sure that the description of the torus, or the CFT
January 23, 2014 28 Ali Nassar
2.1. One-loop partition function
thereof, is independent of the particular choice of the basis. A change of the basis
will take the form (ω′1
ω′2
)=
(a b
c d
)(ω1
ω2
), a, b, c, d ∈ Z. (2.1.7)
The above matrix should be invertible and the requirement that the unit lattice cell
should have the same area in both basis imposes
ab− cd = 1. (2.1.8)
This restricts us to the group SL(2,Z) of 2×2 invertible matrices with integer entries
and unit determinant. The above action on ω1 and ω2 induces the following SL(2,Z)
action on t
t 7→ at+ b
ct+ dwith
(a b
c d
)∈ SL(2,Z), ab− cd = 1 (2.1.9)
which gives an equivalent torus. Since the above action on t remains the same if we
change the signs of a, b, c, d, then only PSL(2,Z) = SL(2,Z)/Z2 acts faithfully. The
group PSL(2,Z) is the modular group of the torus. It can be shown [47], that the
modular group is generated by the following two operations
T : t −→ t+ 1 or T =
(1 1
0 1
)
S : t −→ −1
tor S =
(0 −11 0
).
(2.1.10)
The T and S transformations satisfy
(ST )3 = S2 = I. (2.1.11)
The set of inequivalent tori is parametrized by H+ modulo the T and S transfor-
mations. It can be shown that the fundamental domain of t is
F0 =− 1
2< R(t) ≤ 1
2, ℑ(t) > 0, |t| ≥ 1
. (2.1.12)
January 23, 2014 29 Ali Nassar
2.2. Rational conformal field theory (RCFT)
The values of t ∈ F0 parametrize inequivalent tori, i.e., tori which can’t be trans-
formed into one another by PSL(2,Z). The one-loop partition function defined at a
particular value of t needs to remain the same if we change t by a PSL(2,Z) trans-
formation, i.e., Z(t, t) is a modular group invariant. If the CFT has a Lagrangian
description then one can ensure modular invariance by integrating over F0 with the
correct measure.
2.2 Rational conformal field theory (RCFT)
The number of primary fields which appear in a given CFT will generically be infinite.
However, there are certain classes of CFTs in which the infinite number of primary
fields can reorganize themselves into a finite number of blocks corresponding to the an
extended chiral algebra. These are rational conformal field theories (RCFTs) [46], and
they will be the main subject of this thesis. One arena where RCFTs arise is in string
compactifications on Calabi-Yau manifolds at special values of their moduli, e.g., in
Gepner models [28]. It is important to understand the conditions for rationality. For
example, the simplest compactifications of a string theory are on tori Tm–when are
they described by RCFTs? This question has been studied in [29–32]. In particular,
Wendland [30] derived rationality conditions valid for all torus dimensions m. More
importantly for us, in the case of 2-tori, Gukov and Vafa [32] found a simple, geometric
criterion for rationality. For T 2, the modular parameter τ and the Kahler parameter
ρ must take special values; they must belong to an imaginary quadratic number field.
Such 2-tori have the property of complex multiplication, and are known as CM tori.
For them, a Gauss product exists (see Chapter 3), and was used in [33] to classify the
corresponding RCFTs.
January 23, 2014 30 Ali Nassar
2.2. Rational conformal field theory (RCFT)
The special values of the moduli indicate RCFTs where the infinite number of
fields is organized into a finite set that is primary with respect to an extended chiral
algebra [48]. In the Tm case, the generic boson algebra U(1)m = U(1)k1×· · ·×U(1)kmis extended by vertex operators defined by vectors of the lattice Λm describing Tm ∼=
Rm/Λm. The extended algebra is well understood; it was written explicitly for the
m = 1 case in [48], for example.
To see how a CFT can become rational at special points of its moduli space, we
revisit the simplest example of an RCFT: the c = 1 CFT of a compact free boson on
a circle of rational square raduis R2 = r/s. The radius R plays the role of a geometric
modulus. At a generic value of R, the number of primary fields is infinite. At R2 =
r/s, the chiral algebra of this theory is extended. The infinite list of representations
reconstitute themselves into finitely many irreducible representations of the extended
chiral algebra. In the present case, the chiral algebra is the Kac-Moody algebra
generated by a U(1) current J(z) = ∂X extended by including charged fields E±(z) =
exp[±i√2kX(z)] of dimension k and charge ±2k, where k = rs. The representations
of U(1)k which are local with respect to all vertex operators are labelled by an integer
defined modulo 2k. Defining the even integral lattice Γk = Z√2k, the extended
chiral algebra is the chiral vertex operator algebra A(Γk
)= ⟨J(z), E±(z)⟩. The
representations of the chiral algebra A(Γk) are labeled by a ∈ Γ∗k/Γk
∼= Z/2kZ, where
Γ∗k is the dual lattice of Γk. We define the mode expansions
J(z) =∑n
jnzn−1, E±(z) =
∑n
E±n z
n−k, (2.2.1)
January 23, 2014 31 Ali Nassar
2.2. Rational conformal field theory (RCFT)
The extended chiral algebra in this case is the Moore-Seiberg algebra
[Jn, Jm] = nδm+n,0,
[Jn, E±m] = ±kE±
m+n,[E+
n , E+m
]=[E−
n , E−m
]= 0,[
E+n , E
−m
]=
(m2 − (k − 1)2) · · · (m2 − 1)m
(2k − 1)!δm+n,0 + · · ·
+(√2ki)2k−1
(2k − 1)!
∑q1+···+q2k−1=n+m
: Jq1 · · · J2k−1 : .
(2.2.2)
Another thing to notice about this simplified model is that the chiral algebra depends
only on k while the σ-model modulus in this case is R2 = r/s. Starting from the
algebra U(1)k, there are many candidate rational boson theories, one for each fac-
torization of k into two coprime integers k = rs. The different factorizations of k
give different geometric realizations corresponding to a σ-model on a circle of radius
square R2 = r/s. We will encounter a similar feature when we study the case of Um,K
algebra.
As we mentioned before, an RCFT is specified by a chiral algebra A which is
the Virasoro algebra Vir⊕ Vir or one of its extensions. Examples of extended chiral
algebras are
• Superconformal algebras. These are supersymmetric extensions of the Virasoro
algebra by fermionic currents (of half-integral spins), e.g., the N = 1 and N = 2
superconformal algebras. Space-time supersymmetry in string theory requires
an N = 2 superconformal symmetry on the world-sheet of the superstring [28,
49].
• The semidirect product of the Virasoro algebra with affine Kac-Moody Lie
January 23, 2014 32 Ali Nassar
2.2. Rational conformal field theory (RCFT)
algebras [50–52]. They describe the propagation of strings on group mani-
folds [21, 53].
• W-algebras. These are higher-spin extensions of the Virasoro algebra by bosonic
currents [54,55].
Rationality means that A has a finite set E of irreducible representations Vµ,
µ ∈ E . The partition function in this case is a sesquilinear form with finite number
of terms
Z(t, t) =∑λ,µ∈E
Mλµ χλ(t) χµ(t). (2.2.3)
The positive integers Mλµ express the fact that a representation (λ, µ) of the left
and right copies of A can appear with some multiplicity. Rationality can be stated
in terms ofMλµ: A CFT is rational ifMλµ has a finite rank. We point out that the
the partition function (zero-point function) (2.2.3) of RCFT factorizes into a finite
sum of holomorphic times anti-holomorphic expressions in the modular parameter
of the torus. This holomorphic factorization is a basic property of RCFTs and it
continues to hold for higher-point correlation functions and for RCFTs on higher
genus Riemann surfaces [56–59]. It was shown in [60] that holomorphic factorization
in RCTTs implies that the central charge c and the conformal dimensions h are
rational numbers.
RCFTs are consistently described by a finite set of representations λ ∈ E of a
certain chiral algebra. Moreover the corresponding genus-1 characters χλ(q) form a
finite dimensional unitary representation of the modular group PSL(2,Z)
χλ
(at+ b
ct+ d
)=∑µ∈∈E
Mµλχµ(t), (2.2.4)
where a, b, c, d ∈ Z and ad− bc = 1.
January 23, 2014 33 Ali Nassar
2.3. Lattices
The T and S transformations are represented on the space of characters as
χλ(t+ 1) =∑µ∈∈E
T µλ χµ(t)
χλ
(− 1
t
)=∑µ∈∈E
Sµλχµ(t).
(2.2.5)
The requirement of on-loop modular invariance of the partition function boils down
to the constraints
[T ,M] = [S,M] = 0. (2.2.6)
Every nonnegative integer matrix M with M00 = 1 which satisfies (2.2.6) defines
a physical modular invariant. Classifying the physical modular invariant partition
functions in RCFT amounts to finding all non-negative integer matrices Mλµ with
M00 = 1 subject to (2.2.6). This classification has been completed for the case A(1)1
for all k in [61,62] which gave rise to the ADE pattern. The ADE classification of A(1)1
also leads to the classification of the c < 1 minimal models and their supersymmetric
extension. The other classification which have been found are A(1)2 for all k; A
(1)l , B
(1)l ,
and D(1)l for all k ≤ 3; (A1 ⊕A1)
(1) for all level (k1, k1) (See [63] for references). The
case which is important for us in this thesis is the algebra Um.K = (u(1), · · · , u(1))(1)
for all matrix valued level K [34].
2.3 Lattices
To define a lattice Γ, we start with a vector space V which we take to be Rp,q (that
is, Rp+q with the standard orthonormal basis and a Lorentzian inner product). The
lattice is made of a discrete set of points of the form
Γ =
m=p+q∑i=1
niei |ni ∈ Z
. (2.3.1)
January 23, 2014 34 Ali Nassar
2.3. Lattices
The lattice Γ has a dimension m = p+ q and we assume that ei are linearly indepen-
dent, so that Γ spans Rp+q. In this case, m is the rank of Γ. The signature of Γ is
(p, q).
The metric on Γ, or the intersection form, is given by
Kij = ⟨ei|ej⟩, i, j = 1, . . . ,m, (2.3.2)
where the inner product is computed using the Lorentzian metric on Rp,q. The inter-
section form encodes the lengths of the basis vectors and the angles between them.
The unit cell of the lattice Γ is the set of points
U =m∑i=1
tiei, 0 ≤ ti ≤ 1. (2.3.3)
For example, for a rank-2 lattice spanned by e1 and e2, the unit cell represents the
parallelogram with the sides at 0, e1, e2, e1 + e2. The volume of the unit cell is given
by√detK.
The dual lattice Γ∗ of a lattice Γ is the set of vectors which have integer inner
products with all vectors of Γ
Γ∗ = x ∈ Rp+q | ⟨x|λ⟩ ∈ Z ∀ λ ∈ Γ. (2.3.4)
The basis vectors of Γ∗ will be denoted by e∗i and are defined by
⟨e∗i |ej⟩ = δij. (2.3.5)
The dual lattice Γ∗ is
Γ∗ =
m=p+q∑i=1
nie∗i |ni ∈ Z
. (2.3.6)
The metric or the intersection form of the dual lattice Γ∗ is given by
K∗ij = ⟨e∗i |e∗j⟩ = K−1
ij . (2.3.7)
January 23, 2014 35 Ali Nassar
2.3. Lattices
It follows that the volume of the unit cell of Γ∗ is given by
|Γ∗| = 1√detK
. (2.3.8)
An integral lattice Γ is a lattice for which all the inner products between all lattice
vectors is integer, ⟨λ|λ⟩ ∈ Z. For an integral lattice one can immediately conclude
that Γ ⊆ Γ∗. An integral lattice is even when the products between all lattice vectors
are even integers, ⟨λ|λ⟩ ∈ 2Z. Otherwise the lattice is called odd.
The discriminant of a lattice is the determinant of its intersection matrix, Disc Γ =
det(K). For an integral lattice we have
Disc Γ =
∣∣∣∣Γ∗/Γ
∣∣∣∣ (2.3.9)
A self-dual lattice is one for which Γ∗ = Γ and hence it satisfies Disc Γ = 1, i.e., a
self-dual lattice is unimodular. Even self-dual lattices Γp.q with signature (p, q) can
exist only for
p− q = 0 mod 8 , (2.3.10)
a crucial fact in the construction of the allowed heterotic string theories in 10D, for
example. For p = q, there is a unique even-self dual lattice up to isomorphism.
The intersection form K specifies the lattice Γ up to GL(r,Z) transformations
which preserve the volume of the unit cell or the discriminant. The GL(r,Z) trans-
formations act on the basis of Γ as
e′i = Gijej. (2.3.11)
This action takes the following form on the intersection matrix
K ′ = GTKG. (2.3.12)
January 23, 2014 36 Ali Nassar
2.4. Toroidal compactification
Fixing the discriminant Disc Γ = detK = D of Γ, then K will define an equiva-
lence class of lattices. Two members of the same class are related by the GL(r,Z)
transformation (2.3.12). Similarly, we can define equivalence classes under SL(r,Z)
transformations. These equivalence classes and their construction will be explained
in more detail in Chapter 3.
2.4 Toroidal compactification
In toroidal compactification on an n-torus T n one assumes a space-time metric of the
form
ds2 =d−1∑µ,ν=0
ηµνdXµdXν +
n∑I,J=1
GIJdXIdXJ , (2.4.1)
where D = d + n is the total space-time dimensions and Xµ and Y I represent the
external and internal space-time coordinates, respectively. Here ηµν is the Minkowski
metric and GIJ is the Ricci-flat metric on the torus. The geometry of the torus is
encoded in the constant metric GIJ and in the simplest case the metric takes the form
GIJ = R2IδIJ . (2.4.2)
This is a rectangular torus made up of a product of perpendicular circles with radii
RI , Tn = S1
R1×· · ·S1
Rn. More generally, the metric will have off-diagonal components
corresponding to non-orthogonal circles.
The new feature of strings moving on tori is the existence of winding modes. This
is because one can consider more general periodicity conditions on the embedding
coordinates of the internal space Y I(σ, τ)
Y I(σ + 2π, τ) = Y I(σ, τ) + 2πW I , W I ∈ Z. (2.4.3)
This corresponds to a closed string winding W I times around the Ith circle.
January 23, 2014 37 Ali Nassar
2.4. Toroidal compactification
By turning on a constant background B field3, then the internal part of the action
of the string takes the form
S = − 1
2π
∫d2σ(GIJη
αβ −BIJϵαβ)∂αY
I∂βYJ . (2.4.4)
The left- and right-moving momenta derived from this action are
pIL =W I +GIJ(KJ
2−BJKW
K)
pIR = −W I +GIJ(KJ
2−BJKW
K),
(2.4.5)
where KI ∈ Z and GIJ denotes the inverse metric.
The mass spectrum is given by
1
2M2
0 = (W K)G−1
(W
K
), (2.4.6)
where the 2n× 2n matrix G−1 is
G−1 =
(2(G−BG−1B) BG−1
−G−1B 12G−1
)(2.4.7)
and (W K) is a 2n charge vector.
The mass spectrum is invariant under the the O(n, n,Z) duality group acting as
G −→ AGAT ,
(W
K
)−→
(W ′
K ′
)= A
(W
K
). (2.4.8)
This is the T -duality group of toroidal compactification which generalizes the R →
1/R duality of circle compactification.
The moduli space of toroidal compactification is generated by the n2 truly marginal
operators
ηαβ∂αYI∂βY
J , ϵαβ∂αYI∂βY
J . (2.4.9)
3The B field exists in all oriented string theories.
January 23, 2014 38 Ali Nassar
2.5. Toroidal CFTs
These operators are truly marginal since the model is Gaussian. The n2 couplings
can be organized in a background matrix
E = G+B. (2.4.10)
The only restriction on E is that its symmetric part, G, be positive definite and we
can represent this space of matrices as a coset
M0n,n = O(n, n;R)/[O(n;R)×O(n;R)]. (2.4.11)
To get the physical moduli space, we quotient the above moduli space by the T -duality
group [64]
Mn,n =M0n,n/O(n, n;Z). (2.4.12)
2.5 Toroidal CFTs
The action of the string on an n-torus gives a toroidal CFT with central charge c = n.
The n-torus can be represented as
T n = Rn/Λ, (2.5.1)
where Λ is a rank n lattice.
The chiral algebra of this CFT is the affine U(1)n × U(1)n chiral algebra gener-
ated by the currents j1(z), · · · , jn(z) and j1(z), · · · , jn(z) with the operator product
expansion
jm(z)jn(w) ∼ δmn
(z − w)2. (2.5.2)
The momentum vector pI of a string state in the toroidal compactification on T n
is the charge vector under the gauge group U(1)n × U(1)n
pI = (pIL, pIR). (2.5.3)
January 23, 2014 39 Ali Nassar
2.5. Toroidal CFTs
This charge vector lives on a charge lattice Γn,n with signature (n, n) and with a scalar
product given by
p2 = p2L − p2R. (2.5.4)
One-loop modular invariance forces the lattice Γn,n to be an even self-dual lattice.
The charge lattice uniquely determines the toroidal CFT and the moduli space of
toroidal CFTs (2.4.12) is equivalent to the moduli space of even self-dual lattices
[23,24].
The set of primary fields up of the vertex operators
Vp(z, z) =: eipX(z, z) : . (2.5.5)
The energy momentum tensor is given by the Sugwara construction
T (z) =1
2
n∑i=1
: jiji : (z), T (z) =1
2
n∑i=1
: jiji : (z). (2.5.6)
Thus the conformal dimension of Vp(z, z) is given by
(h, h) =(p2L2,p2R2
). (2.5.7)
The primary fields have simple fusions in which their charges add up. The operator
product expansion of the primary fields takes the form
Vp(z, w)× Vp′(z, w) ∼ cpp′′(z − w)pLp′L(z − w)pRp′RVp+p′(w, w). (2.5.8)
On the circle (z − w)(z − w) = 1, the above OPE takes the form
Vp(z, w)× Vp′(z, w) ∼ cpp′(z − w)pLp′L−pRp′RVp+p′(w, w) (2.5.9)
which defines the natural metric on the lattice of charges Γn,n to be
p p′ = pLp′L − pRp′R. (2.5.10)
January 23, 2014 40 Ali Nassar
2.5. Toroidal CFTs
The partition function is
ZΓn,n(t, t) =1
|η|2n∑
(p,p)∈Γn,n
q12p2L q
12p2R , (2.5.11)
where q = exp(2πit) and the Dedkind η function is defined as
η = q24∞∏n=1
(1− qn). (2.5.12)
The partition function is modular invariant by the even self-duality of the lattice Γn,n.
The charge lattice or the Narain lattice of the toroidal CFT is given by
Γ(G,B) =(pL; pR
):=
1√2
(µ− Bλ+ λ;µ− Bλ− λ
)| (µ, λ) ∈ Λ∗ ⊕ Λ
, (2.5.13)
where Λ∗ and Λ are the momentum and winding lattices, respectively and B = ΛT BΛ,
where Λ also denotes the intersection form of the lattice Λ. The metric is given in
terms of Λ as
G = ΛTΛ, (2.5.14)
The holomorphic and anti-holomorphic vertex operators are characterized by pR =
0 and pL = 0, respectively. They are parametrized by the values of their charges in
ΓL = (pL; 0) and ΓR = (0; pR)
ΓL =(pL; 0
):=
1√2
(µ− Bλ+ λ; 0
)|(µ, λ) ∈ Λ∗ ⊕ Λ
ΓR =
(0; pR
):=
1√2
(0;µ− Bλ− λ
)|(µ, λ) ∈ Λ∗ ⊕ Λ
.
(2.5.15)
We also define the following projections of the lattice Γ(τ, ρ):
ΓL = (pL; ∗) :=1√2
(µ− Bλ+ λ; ∗
)|(µ, λ) ∈ Λ∗ ⊕ Λ
ΓR = (∗; pR) :=
1√2
(∗;µ− Bλ− λ
)|(µ, λ) ∈ Λ∗ ⊕ Λ
,
(2.5.16)
January 23, 2014 41 Ali Nassar
2.5. Toroidal CFTs
where pR = ∗ means pR can take any value and doesn’t constraint pL. It can be
shown that (see (2.7.6))
ΓL = Γ∗L. (2.5.17)
For a generic point in the Narain moduli space (2.4.12), the number of primary
fields is infinite and the toroidal CFT is not rational. However, for particular choices
of the target space data G and B, the chiral algebra is extended by additional holo-
morphic vertex operators. With respect to this newW-algebra, the infinite number of
primary fields can be arranged into a finite number of blocks and one gets a rational
CFT.
The conditions for rationality of a 2-torus CFT were analyzed in detail in [32,33].
This is based on previous work by Moore [31] and Moore et al. [29]. Their conclusion
is that a Narain lattice Γ(τ, ρ) is rational iff the following equivalent conditions are
satisfied
(1) rank(ΓL
)= rank
(ΓR
)= 2.
(2) τ, ρ ∈ Q(D) for some integer D < 0.
In [32], it was shown that the above conditions are equivalent to the following
statement
• A Narain lattice Γ(τ, ρ) is rational iff the lattice ΓL is a finite-index sublattice
of ΓL.
In this case the number of primary fields of the rational theory or the dimension of
the chiral ring is given by the index
|D| = [ΓL : ΓL] (2.5.18)
January 23, 2014 42 Ali Nassar
2.6. c = 1
Using (2.5.17), the group ΓL/ΓL = Γ∗L/ΓL is the discriminant group of ΓL [65]. It is a
finite Abelian group of order |D| and another way of writing the number of primary
fields is
|D| = |Γ∗L/ΓL|. (2.5.19)
The generalization to higher dimensional tori has been done in [30] where it was
shown that if C(G,B) is a toroidal conformal field theory with target space data G and
B and with a central charge c = m. Then C(G,B) is rational iff G := ΛTΛ ∈ GL(d,Q)
and B ∈ Skew(d)∩Mat(d,Q). The W-algebra of C(G,B) is generated by the n U(1)
currents besides the extra holomorphic vertex operatorsWλiwhich correspond to the
generators λi of the lattice ΓL.
It was shown in [30,31] that the points in Narain moduli space which correspond to
RCFTs are dense. It is argued in [31] that a similar density property holds for RCFTs
corresponding to string compactification on K3 surfaces. The generalization to the
Calabi-Yau three-folds has been done in [32] where a criterion for the rationality
of σ-models on Calabi-Yau three-folds has been conjectured. The authors of [32]
conjectured that a CFT based on Calabi-Yau three-fold M is rational iff M and
its mirror W admit complex multiplication over the same number field. Complex
multiplication for Calabi-Yau manifolds will be explained in the next chapter.
2.6 c = 1
In this section we study the simplest example of a compact CFT of a real massless
scalar field X which lives on a circle of radius R, i.e., X ∼ X+2πR. This is the c = 1
CFT of a compact free boson on a circle of radius R ∈ R≥0. The radius is the only
geometric modulus in this model. The T -duality acts on R as t : R ↔ 1/R (we are
January 23, 2014 43 Ali Nassar
2.6. c = 1
working in units where α′ = 2) and gives an isomorphic CFT and the moduli space
of CFTs is given by
M = R+/t. (2.6.1)
The action is
S =
∫d2x[∂1X∂1X + ∂2X∂2X
](2.6.2)
In complex coordinates, the action takes the following form
S =
∫dz dz ∂X(z, z)∂X(z, z). (2.6.3)
The equation of motion is
∂∂X(z, z) = 0 (2.6.4)
implying that
j(z) = i∂X(z, z) holomorphic field
j(z) = i∂X(z, z) antiholomorphic field.
(2.6.5)
where j(z) is the chiral current which generates the algebra U(1).
The energy momentum tensor is
T (z) = −1
2∂X∂X =
1
2j(z)j(z), T (z) = −1
2∂X∂X =
1
2j(z)j(z). (2.6.6)
The left-moving and the right moving momenta are given by
pl =1√2
( nR
+mR), pr =
1√2
( nR−mR
), m, n ∈ Z. (2.6.7)
The vector (pL, pR) can be expressed as
(pL, pR) = me1 + ne2, (2.6.8)
January 23, 2014 44 Ali Nassar
2.6. c = 1
where
e1 =1√2
( 1
R,1
R
), e2 =
1√2(R,−R). (2.6.9)
With respect to the inner product (2.5.10), we have e1 · e1 = e2 · e2 = 0 and e1 · e2 =
e2 · e1 = 1. The intersection form of the lattice Γ1,1 spanned by e1 and e2 is
K =
(0 1
1 0
). (2.6.10)
The lattice Γ1,1 is an even self-dual lattice. This is the momentum-winding Narain
lattice in which (pL, pR) lives.
The partition function is
ZΓ1,1(t, t;R) =1
|η|2∑
(p,p)∈Γ1,1
q12p2L q
12p2L , (2.6.11)
For generic values of R inM, the models are not rational and the chiral algebra is
the U(1) algebra generated by the current j(z). One can try to find extensions of the
chiral algebra by looking for new holomorphic vertex operators which might extend
it. A holomorphic vertex operator Vγ(z) will appear at pR = 0. The new holomorphic
vertex operators will have the form
Vγ(z) = eiγX(z). (2.6.12)
These are the only vertex operators allowed in the free boson theory. The conformal
dimension of Vγ(z) is γ2/2.
Looking at the second equation in (2.6.7), the constraint pR = 0 gives
R2 =n
m(2.6.13)
which makes sense only when R2 is rational, i.e., holomorphic vertex operators will
appear only when R2 is rational. We will put R2 = r/s, where r and s are two coprime
January 23, 2014 45 Ali Nassar
2.6. c = 1
integers. The T duality R↔ 1/R now acts as r ↔ s. The conformal dimension must
be invariant under r ↔ s. Since X is periodic, then eiγX(z) must be well defined
under
X → X + 2π
√r
s. (2.6.14)
The above requirements forces γ to take the form
γ = ±√2rs (2.6.15)
where the factor of 2 is included to give Vγ(z) an integer conformal dimension.
The extended chiral algebra is generated by
j(z) = i∂X(z), V±√2rs = ei±
√2rsX(z). (2.6.16)
For r = s = 1, this gives the affine su(2)1 extension of the free boson theory at the
self-dual radius R = 1.
The primary fields are given by the vertex operators
Φα(z, z) = eiαX(z, z). (2.6.17)
Since the operator product expansion of V±√2rs and Φα(z, z) can’t have branch cuts,
the value of α is fixed by locality to be
α =l√2rs
, l ∈ Z. (2.6.18)
The periodicity of X can be used to fix the fundamental domain of l to
l = −2rs+ 1,−2rs+ 2, · · · , 2rs. (2.6.19)
Let us formulate the above results in terms of even integral lattices. First we define
the lattice
ΓL = Ze = Z√2rs. (2.6.20)
January 23, 2014 46 Ali Nassar
2.6. c = 1
Since ⟨e|e⟩ = 2rs ∈ 2Z, then the lattice ΓL is even integral. We will call it the lattice
of holomorphic vertex operators. The dual lattice Γ∗L is given by
Γ∗L = Ze∗ =
Z√2rs
. (2.6.21)
Looking at the weights of the primary fields (2.6.18) and their fundamental domain
(2.6.19), one learns that the set of primary fields are parametrized by elements of the
coset (or the discriminant group)
Γ∗L/ΓL. (2.6.22)
The dimension of this coset is nothing but the determinant of the intersection matrix
of the lattice ΓL
D = 2rs. (2.6.23)
This is the simplest realization of (2.5.19) which gives the dimensions of the chiral
ring or the number of primary fields in the terms of the discriminant of the lattice of
the holomorphic vertex operators.
In the language of the Lie algebra, the above results can be phrased in the following
way. The lattice Γ∗L is the weight lattice and the lattice ΓL is the root lattice. This
is so since the action of a holomorphic vertex operator, parametrized by an element
in ΓL, shifts the weight of the primary fields by an element of Γ∗L. The ratio Γ∗
L/ΓL
has |ΓL| congruence classes which labels the set of admissible representations.
The above discussion shows that the U(1) chiral algebra is extended at rational
values of R2 with the appearance of new holomorphic vertex operators. The require-
ment of locality with respect to the new holomorphic vertex operators selects only a
finite number of primary fields and the theory becomes rational. Now let us see how
can we formulate the above discussion in terms of the properties of the charge lattice
Γ1,1.
January 23, 2014 47 Ali Nassar
2.6. c = 1
First, let us study the projections of Γ1,1 at R2 = r/s. The left-moving momentum
lattice which is spanned by the vectors pL
ΓL ≡1√2
( nR
+mR), m, n ∈ Z (2.6.24)
is a rank-2 lattice for generic R. However, for R2 = r/s, one gets
ΓL =Z√2rs
. (2.6.25)
This equation says that the vectors pL live in an even integer lattice ΓL of rank 1
which is spanned by e = 1/√2rs. This is the weight lattice which we found earlier
(2.6.18). The dual lattice Γ∗L of ΓL is given by
Γ∗L = Ze∗ = Z
√2rs. (2.6.26)
This the lattice of holomorphic vertex operators (2.6.20) which is a rank-1 lattice.
The fact that
Γ∗L = ΓL −→ ΓL = Γ∗
L (2.6.27)
is easy to understand in terms of the projections Γ1,1. First, since ΓL parameterizes
the holomorphic vertex operators then its vector will have the form
p = (pL, 0). (2.6.28)
The lattice ΓL being the lattice of the left-moving momentum its vectors will have
the form
p′ = (p′L, ∗). (2.6.29)
Now since
p p′ = pLp′L ∈ Z. (2.6.30)
January 23, 2014 48 Ali Nassar
2.7. c = 2
Then (2.6.27) follows.
The number of primary fields, or the dimension of the chiral ring, now follow easily
|D| = |Γ∗L/ΓL| = |ΓL/ΓL| (2.6.31)
This number is finite precisely because ΓL is a finite-index sublattice of ΓL which is
the condition for rationality alluded to earlier.
2.7 c = 2
From the simple example of the previous section, we learn that rationality emerges
when ΓL is a finite index sublattice of ΓL and in this case, the number of primary
fields or the dimensions of the chiral ring is given by |Γ∗L/ΓL|. In this section, we
review the analogous construction for strings on a 2-torus [32]. Rationality can be
translated into a simple, geometric condition on the modular parameter τ and the
Kahler parameter ρ of the 2-torus.
We consider a 2-torus, or an elliptic curve, with a metric
ds2 = G11dx2 + 2G12dxdy +G22dy
2
=ρ2τ2
∣∣dx+ τdy∣∣2. (2.7.1)
The complex structure modulus and the Kahler modulus are given by
τ = τ1 + iτ2 =G12
G11
+ i
√detG
G11
,
ρ = ρ1 + iρ2 = B + i√detG.
(2.7.2)
The momentum-winding Narain lattice Γ2,2 contains(pL
pR
)=
i√2τ2ρ2
Z
(1
1
)⊕
(ρ
ρ
)⊕
(τ
τ
)⊕
(ρτ
ρτ
). (2.7.3)
January 23, 2014 49 Ali Nassar
2.7. c = 2
We define the following projections of the lattice Γ2,2 (here we consider only the
left-moving part of Γ2,2 but all the definition will apply analogously to the right-
moving part)
ΓL =pL | pR = 0,
(pL; 0
)∈ Γ2,2
ΓL =
pL | pR = ∗,
(pL; pR
)∈ Γ2,2
.
(2.7.4)
It is clear that ΓL ⊆ ΓL. Also, since
(pL, 0) (qL, ∗) = pLqL ∈ Z (2.7.5)
Then, ΓL and ΓL are dual lattices
ΓL∼= Γ∗
L. (2.7.6)
The lattice ΓL will in general have a rank zero since the constraint pL = 0 will
have no solutions except for special values of the moduli τ and ρ. For these special
values of τ and ρ, the lattice ΓL will become a rank-2 sublattice of Γ2,2. Looking at
the explicit basis of the Narain lattice (2.7.3 ), and since ΓL is rank-2, then it is cut
by two independent equations. Since ΓL corresponds to pR = 0, the two equations
have the form
m1 +m2ρ+m3τ +m4τρ = 0,
m′1 +m′
2ρ+m′3τ +m′
4τρ = 0.(2.7.7)
By solving for ρ from the second equation and plugging the result into the first, we
arrive at
aτ 2 + bτ + c = 0, a, b, c ∈ Z gcd(a, b, c) = 1. (2.7.8)
This implies that
τ =−b+
√D
2a, D = b2 − 4ac < 0, (2.7.9)
January 23, 2014 50 Ali Nassar
2.8. Matrix-level affine Kac-Moody algebra
where we have chosen the solution which gives τ a positive imaginary part. Substitut-
ing this value of τ back into the first equation in (2.7.7), we learn that ρ also satisfies
a quadratic equation with the same discriminant D. Hence both τ and ρ belong to
an imaginary quadratic number field Q(√D).
The criteria for rationality found in [32] for a CFT based on an elliptic curve E is
RCFT⇐⇒ τ, ρ ∈ Q(√D). (2.7.10)
Elliptic curves for such values of τ, ρ are said to have complex multiplication or to be
of CM type. We will study them in detail in the next chapter.
2.8 Matrix-level affine Kac-Moody algebra
In this section we introduce the matrix-level affine Kac-Moody algebras based on a
group G. The Um,K algebra will arise as a special case when G = U(1). Let G be a
compact, simple, semi-simple Lie group with a Lie algebra g. The affine Kac-Moody
algebra of g is gk, where k ∈ Z≥0 is the level. We study the affine extension of
g⊕m with matrix-valued level KAB. We take KAB to be the intersection form of a
lattice ΓK . The inverse KAB is the intersection form of the dual lattice Γ∗K . We will
denote the currents by JaA, where a is the Lie algebra index a = 1, · · · , dim(g) and
A = 1, · · · , N = rank(K). The Lie algebra of g⊕m takes the form[JaA, J
bB
]= ifabcJ
cBδAB. (2.8.1)
We assume the following OPE of the currents
JaA(z)J
bB(w) ∼
ifabcJcB(w)δAB
z − w+
δabKAB
(z − w)2. (2.8.2)
This is not the central extension of the loop group of g⊕m since a cohomological
obstruction prevents the construction of a two-cycle on the loop group of g⊕m. This
January 23, 2014 51 Ali Nassar
2.8. Matrix-level affine Kac-Moody algebra
however doesn’t mean that one can’t construct a CFT corresponding to the above
OPE.
We use the Sugawara construction to find the energy-momentum tensor
T (z) = GAB(JaAJ
aB)(z), (2.8.3)
where GAB is such that the currents JaA(z) have conformal dimension one
T (z)JaA(w) ∼
JaA(w)
(z − w)2. (2.8.4)
The matrix GAB turns out to be
GAB =KAB
2(hK + 1), (2.8.5)
where KAB = K−1AB, h is the dual Coxeter number of the group G and K = Tr(KAB).
Using the energy-momentum tensor
T (z) =KAB
2(hK + 1)(Ja
AJaB)(z) (2.8.6)
we can find the central charge from the most singular term of the OPE of T (z) with
itself
T (z)T (w) ∼ c/2
(z − w)4+ · · · (2.8.7)
The value of c is
c(K) =N dim g
hK + 1. (2.8.8)
For N = 1 and K = k−1, we recover the usual value of c
c(k) =k dim g
h+ k. (2.8.9)
On the other hand, in the diagonal case KAB = kAδAB, the central charge is not a
sum
c(K) =∑A
c(kA). (2.8.10)
January 23, 2014 52 Ali Nassar
2.8. Matrix-level affine Kac-Moody algebra
The Abelian case corresponds to the algebra u(1) ⊕ · · · ⊕ u(1) which we denote
by Um,K , where K is an m×m matrix-valued level. This algebra was studied in [34]
where its modular invariant partition functions were classified. The operator product
expansion of the currents takes the form
JA(z)JB(w) ∼KAB
(z − w)2. (2.8.11)
The motivation for studying Kac-Moody algebras with matrix-valued level is the
existence of Chern-Simons theories on a three-manifold N3, based on a gauge group
G = U(1)m with the level being an integer valued matrix [66]
S = KAB
∫M4
FA ∧ FB, (2.8.12)
where M4 is the four manifold for which N3 is a boundary. The independence
of the partition function from the extension from N3 to M4 forces KAB to be an
integer-valued matrix. The existence of the above action relies on the fact that
H4(BU(1),Z) = Z, where BU(1) is the classifying space of U(1). For G = U(1)m,
BG = Zm.
The study of the matrix-level affine algebra Um,K is partly motivated by the ef-
fective description of the fractional quantum Hall effect, via a Chern-Simons theory
with a gauge group U(1)m and a matrix valued level K [66, 67]. The Witten corre-
spondence [48,68] relates the Chern-Simons theory with a gauge group G canonically
quantized on a manifold M = Σ× Rt, to the RCFT based on the affine Kac-Moody
algebra Gk (the Wess-Zumino-Witten model). For any simple, compact gauge group
G the Chern-Simons actions on a three-manifold are classified by an integer k ∈ Z.4
It was shown in [66] that Chern-Simons theories for the Abelian group U(1)m are
4If the three-manifold is a spin manifold then k can be half-integer [69].
January 23, 2014 53 Ali Nassar
2.8. Matrix-level affine Kac-Moody algebra
classified by positive integer lattices with intersection forms K. Due to their poten-
tial application in the fractional quantum Hall effect, it is of interest to study the
two-dimensional avatars of the U(1)m Chern-Simons theories based on matrix val-
ued level K. This reduces to the study of two-dimensional RCFTs with the affine
Kac-Moody algebra Um,K .
Chern-Simons theories with matrix-valued level K gives an effective description of
the Abelian Fractional Quantum Hall (FQH) liquids [66,67,70,71]. The matrix-valued
level K appears in a class of multi-layer FQH states which generalizes the Laughlin
state [72]. The wave function of those FQH multi-layer states takes the form
Ψ(z) =∏a,b,i,j
(zai − zbj)Kab/2 exp(−∑|zai|2
), (2.8.13)
where zai is the coordinate of the ith electron in the ath layer. The matrix K deter-
mines the patterns of the zeros of the wave function and the various monodromies
which occur when electrons moves around each other. This can be related to the
number of the flux quanta attached to the electrons in each layer. For example, a
zero of order Kaa/2 in the above wave function means the phase of the wave function
will acquire a phase Kaaπ when the electrons circle one another. This in return means
there are Kaa/2 flux quanta attached to each electron and the matrix K determines
how flux quanta are attached to the electrons.
Since the gauge group of the action (2.8.12) is compact then the charges which
appear in the theory are quantized and they live in a charge lattice. The allowed
field redefinitions on the field strength FA in (2.8.12) are the ones which preserve
the integrality of the charges. These field transformations are elements of the group
SL(m,Z) and they act on the matrix K as
K −→ SKST , (2.8.14)
January 23, 2014 54 Ali Nassar
2.8. Matrix-level affine Kac-Moody algebra
where S ∈ SL(m,Z).
Two FQH states (2.8.13) with their K matrices related by the SL(m,Z) transfor-
mation (2.8.14) are considered equivalent, i.e., they belong to the same universality
class. Hence, multi-layer FQH states are classified by SL(m,Z) equivalence classes
of K matrices where the equivalence relation is (2.8.14). This fact will be used later
to define a new product on the set of FQH states (see the discussion after (5.0.5)).
January 23, 2014 55 Ali Nassar
Chapter 3
Complex multiplication
3.1 Elliptic curves
In this chapter we describe in more detail complex multiplication for elliptic curves,
higher-dimensional tori, and for Calabi-Yau manifold. Our aim is to set the stage for
the conjecture by Gukov and Vafa which relates the rationality of σ-models on Calabi-
Yau manifolds to the complex multiplication property of these manifolds. We will give
a physicist-friendly introduction to some concepts in algebraic number theory and the
theory of elliptic curves which will be important later. We will follow [31,47,73].
A elliptic curve is a genus-1 surface. When viewed as real manifolds, all elliptic
curves are diffeomorphic to a product of two circles S1 × S1. However, when they
are given a complex structure, i.e., when viewed as complex manifolds, elliptic curves
are not all equivalent. Actually, there is an infinite number of elliptic curves with
different complex structures, as we will see later.
A rigorous definition of an elliptic curve requires some technical tools from al-
gebraic geometry. We will start with the definition of an elliptic curve using the
algebraic equation:
f(x, y) = y2 − 4x3 + g2x+ g3 = 0, x, y ∈ C, (3.1.1)
56
3.1. Elliptic curves
where g2 and g3 are complex numbers. This is the Weierstrass equation of the elliptic
curve. Later on we will relate this definition to complex tori.
In order for the elliptic curve (3.1.1) to be well-defined one needs to make sure it is
non-singular (no cusps, no self-intersection, no isolated points). The curve will be non-
singular iff there are no points for which f(x, y) = 0 and df(x, y) = 0 simultaneously.
For (3.1.1), this translates to
∆ = 4g32 + 27g23 = 0. (3.1.2)
The polynomial ∆ is known as the discriminant of the elliptic curve.
A complex torus T 2 can be constructed in the following way: start with a lattice
Λ in the complex plane
Λ = Zω1 ⊕ Zω2, (3.1.3)
where ω1, ω2 are the basis vectors of the lattice. Then a complex torus is the
quotient of the complex plane by the lattice Λ
C/Λ = z ∼ z + Λ : z ∈ C. (3.1.4)
This definition makes it obvious that the complex torus is an Abelian group under
addition since both C and Λ are Abelian groups. As such, a complex torus is an
example of an Abelian variety. It is also a Riemann surface of genus one.
By rescaling we can bring the lattice to the form
Λ = m1 + nτ, τ =ω1
ω2
∈ H+, (3.1.5)
where H+ is the upper-half of the complex plane. The holomorphic coordinate on the
torus is
z = x+ τy. (3.1.6)
January 23, 2014 57 Ali Nassar
3.1. Elliptic curves
As we explained before in Chapter 2, the SL(2,Z) action on ω1 and ω2 induces
the following SL(2,Z) action on τ
τ 7→ aτ + b
cτ + dwith
(a b
c d
)∈ SL(2,Z), ab− cd = 1 (3.1.7)
which gives an equivalent torus. Since the above action on τ remains the same if we
change the signs of a, b, c, d, then only PSL(2,Z) = SL(2,Z)/Z2 acts faithfully. The
group PSL(2,Z) is the modular group of the torus. The modular group is generated
by the two operations [47]
T : τ −→ τ + 1 or T =
(1 1
0 1
)
S : τ −→ −1
τor S =
(0 −11 0
).
(3.1.8)
The T and S transformations satisfy
(ST )3 = S2 = 1. (3.1.9)
The set of inequivalent tori is parametrized by H+ modulo the T and S transfor-
mations. The fundamental domain of τ is [47]
F0 =− 1
2< R(τ) ≤ 1
2, ℑ(τ) > 0, |τ | ≥ 1
. (3.1.10)
The values of τ ∈ F0 parametrize inequivalent tori, i.e., tori which can’t be trans-
formed into one another by PSL(2,Z).
An elliptic curve is isomorphic to a complex torus. To establish this isomorphism
we need to define the Weierstrass function. First, an elliptic function is a function on
C which is doubly periodic for Λ
F (z + ω1) = F (z)
F (z + ω2) = F (z).(3.1.11)
January 23, 2014 58 Ali Nassar
3.1. Elliptic curves
The Weierstrass function is an elliptic function defined as
℘(z) =1
z2+∑ω∈Λ
′[ 1
(z − ω)2− 1
ω2
]. (3.1.12)
where∑′ means we only sum over non-zero lattice vectors. Using the notation
sm(Λ) =∑ω∈Λ
′ 1
ωm(3.1.13)
we get the expansion
℘(z) =1
z2+
∞∑n=1
(2n+ 1)s2n+2(Λ)z2n. (3.1.14)
The first few terms are
℘(z) =1
z2+ 3s4(Λ)z
2 ++5s6(Λ)z4. (3.1.15)
Now we state without proof the following theorem [47].
Theorem 3.1.1. Let g2 = g2(Λ) = 60s4 and g3 = g3(Λ) = 140s6. Then
℘′2 = 4℘3 − g2℘− g3. (3.1.16)
This means that the points (x, y) = (℘(z), ℘′(z)) lie on the elliptic curve
y2 = 4x3 − g2x− g3. (3.1.17)
It can be shown that the discriminant of this elliptic curve ∆(Λ) doesn’t vanish and
the equation defines a non-singular elliptic curve.
The j-invariant of the lattice Λ is defined as
j(Λ) = 1728g2(Λ)
g2(Λ)3 − 27g3(Λ)2= 1728
g2(Λ)
∆(Λ)(3.1.18)
This is always well defined since ∆(Λ) = 0.
January 23, 2014 59 Ali Nassar
3.1. Elliptic curves
One of the properties of the j-invariant is that it is invariant under scaling of the
lattice Λ. Two lattices Λ and Λ′ are said to be homothetic if there is a non-zero
complex number λ such that
Λ′ = λΛ (3.1.19)
For example any lattice Λ = Zω1 ⊕ Zω2 is homothetic to a lattice of the form Λ =
Z⊕Zτ . Since we are only interested in lattices up to a homothety, then the j-invariant
is useful because it is defined over the equivalence classes of homothetic lattices
j(Λ) = j(λΛ). (3.1.20)
This can be easily proven by noting that
g2(λΛ) = λ−4g2(Λ)
g3(λΛ) = λ−6g6(Λ).(3.1.21)
When defined for the lattice Λτ = Z ⊕ Zτ , the j-invariant becomes j(τ). The
function j(τ) is invariant under the SL(2,Z) group
j
(aτ + b
cτ + d
)= j(τ). (3.1.22)
In particular
j(τ + 1) = j(τ). (3.1.23)
This implies that j(τ) is holomorphic in q = e2πiτ , where 0 < |q| < 1 since ℑ(τ) > 0.
The j function has the following q-expansion [47]
j(τ) =1
q+
∞∑n=0
cnqn =
1
q+ 744 + 196884q + · · · , (3.1.24)
where τ ∈ H and cn ∈ Z∀n ≥ 0.
Now we state one of the most important theorems in the study of elliptic curves.
January 23, 2014 60 Ali Nassar
3.1. Elliptic curves
Theorem 3.1.2 (Uniformization theorem). Let E be an elliptic curve given by theWeierstrass equation
y2 = 4x3 − g2x− g3, (3.1.25)
where g2, g3 ∈ C and g32 − 27g23 = 0. Then there is a unique lattice Λ such that
g2 = g2(Λ)
g3 = g3(Λ)(3.1.26)
The isomorphism between complex tori and elliptic curves can be written as
ψ : C/(Z+ τZ) −→ E
z 7→ ψ(z) = [℘(z), ℘′(z)].(3.1.27)
An elliptic curve is also a one-dimensional Calabi-Yau manifold. It can be repre-
sented by a homogeneous degree-3 polynomial in CP2
a1Z31 + a2Z
32 + a3Z
33 + · · ·+ a10Z2Z
23 = 0. (3.1.28)
This is the natural description of the elliptic curve as a target space in string theory
in terms of a gauged linear sigma model [74].
There are 10 parameters ai in (3.1.28) corresponding to the 10 independent degree-
3 monomials in 3 variables. One parameter can be removed by an overall scaling and 8
parameters can be removed by GL(3,C) transformations. The remaining parameter
characterizes the complex structure of the elliptic curve, i.e., the moduli space of
complex structures on an elliptic curve is one-dimensional. We can go from the
projective representation (3.1.28) to the Weierstrass form (3.1.1) by first rescaling
out one of the projective coordinates, say Z1. Then the Weierstrass form can be
obtained by coordinate redefinitions of the remaining coordinates. The g2 and g3
which appear in the Weierstrass form will be functions of the one parameter which
remains in (3.1.28). The projective form of the Weierstrass equation is
zy2 = 4x3 + g2z2x+ z3g3 (3.1.29)
January 23, 2014 61 Ali Nassar
3.1. Elliptic curves
which reduces to the affine form once we scale z out using the projective equivalence
(x, y, z) ∼ (βx, βy, βz), where β is a non-zero complex number.
The j-invariant allows one to read the value of τ from the projective form of the
elliptic curve. First, we reduce equation (3.1.28) to the Weierstrass form (character-
ized by g2 and g3 which are now functions of the parameters ai) using GL(3,C). Then
τ can be found from
j(τ) = 1728g2(Λ)
∆(Λ). (3.1.30)
For example, consider the elliptic curve represented by the Fermat cubic
Z31 + Z3
2 + Z33 = 0. (3.1.31)
Using the change of coordinates
Z1 =1
6z +
√3
18y, Z2 =
1
6z −√3
18y, Z3 = −
1
3x (3.1.32)
we get
zy2 = 4x3 − z3. (3.1.33)
This means g2 = 0 and g3 = 1 hence j(τ) = 0. This value of j corresponds to
τ = eiπ/3. We we will see in the next section that elliptic curves with this value of τ
have complex multiplication.
The complex torus admits a nowhere-vanishing holomorphic 1-form ω = dz. The
integrals of ω around the two periods of the torus A,B define the period
τ =
∫Bω∫
Aω. (3.1.34)
It is more convenient to normalize ω such that∫Aω = 1 and as such
τ =
∫B
ω. (3.1.35)
January 23, 2014 62 Ali Nassar
3.1. Elliptic curves
The periods A,B are the basis of 1-cycles on the torus (a genus-1 surface). There is
an SL(2,Z) ambiguity in the choice of the basis of 1-cycles which translates to the
SL(2,Z) action on τ .
An n-dimensional complex torus T 2n with complex coordinates z1, · · · , zn is de-
scribed using the identification
zi ∼ zi + δij, zi ∼ zi + Tij, (3.1.36)
where the n × n period matrix Tij satisfies ℑ(Tij) > 0. The column vectors of Tij
define a rank-n lattice Λ. The torus T n is also given as a quotient T 2n = Cn/Λ.
To define the notion of complex multiplication for Calabi-Yau manifolds, we first
need to define the Jacobian of a genus-n Riemann surface Σn. On Σn there are n
holomorphic 1-forms ωi and 2n 1-cycles Ai, Bi, i = 1, . . . , n with a symplectic pairing
Ai ∩Bj = δij, Ai ∩ Aj = Bi ∩Bj = 0. (3.1.37)
The holomorphic 1-forms ωi are normalized relative to the Ai-cycles such that∫Ai
ωj = δij. (3.1.38)
The period matrix is defined as
Tij =∫Bi
ωj. (3.1.39)
There is an SL(2n,Z) ambiguity in the choice of the basis of 1-cycles which translates
to the SL(2n,Z) action on T .
For Calabi-Yau manifolds, there is a simple formula for the period matrix TIJ
in terms of the prepotential F of the complex-structure moduli space. To define
F , we first describe the complex-structure moduli space in more detail. We look at
January 23, 2014 63 Ali Nassar
3.2. Quadratic number fields
the middle-homology cycles which for a Calabi-Yau three fold X are three (real)-
dimensional and they live in H3(X,Z). The dimension of H3(X,Z) is given by the
third Betti number b3(X). For a Calabi-Yau X, b3(X) can be refined in terms of the
dimensions of the Dolbeault cohomology groups as
b3(X) = h2,1 + h1,2 + h3,0 + h0,3 = 2h2,1 + 2. (3.1.40)
A basis of three-cycles AI , BJ , I, J = 0, · · · , h2,1 can be chosen such that
AI ∩BJ = δIJ , AI ∩ AJ = BI ∩BJ = 0. (3.1.41)
To define a set of coordinates on the complex-structure moduli space, we need a
quantity which depends on the complex structure. This is provided by the unique
nowhere-vanishing holomorphic 3-form Ω which exists on any Calabi-Yau manifold.
The integrals of Ω over A and B give
XI =
∫AI
Ω, FI =
∫AI
Ω, I = 0, · · · , h2,1. (3.1.42)
Since the dimensions of the complex-structure moduli space is h2,1, the above coor-
dinates are overcomplete and in fact, FI are functions of XI .
The prepotential of the Calabi-Yau manifold is defined as
F =1
2XIFI(X
I). (3.1.43)
The period matrix is now given by
TIJ =F
∂XI∂XJ. (3.1.44)
3.2 Quadratic number fields
Let Q(θ) denote the set of numbers of the form
x = x1 + x2θ (3.2.1)
January 23, 2014 64 Ali Nassar
3.3. Quadratic number fields
where x1, x2 ∈ Q. The set Q[x] will denote the ring of polynomials in x with co-
efficients from Q. While the set Z[x] will denote the ring of polynomials in x with
coefficients from Z.
Definition 3.2.1. An algebraic number α is a complex number α ∈ C which satisfiesa polynomial equation with rational coefficients
f(α) = 0, f ∈ Q[x]. (3.2.2)
For example,√2, i are algebraic numbers since they are solutions of x2 − 2 = 0
and x2 + 1 = 0, respectively.
If the polynomial f(α) = 0 has integer coefficients, i.e., f ∈ Z[x] where f is degree
m polynomial together with am = 1, then α is an algebraic integer.
Definition 3.2.2. An algebraic number field is a subfield L of C such that [L : Q] <∞, where [L : Q] is the dimension of L when viewed as a vector space over Q. Thenumber [L : Q] is called the degree of L over Q.
Theorem 3.2.1. If L is a number field then L = Q(α) for some algebraic number α.
Definition 3.2.3. An algebraic number field L is quadratic if [L : Q] = 2. Hence,there is a number θ such L = Q(θ), where
θ =−b±
√b2 − 4ac
2a, (3.2.3)
i.e., θ satisfiesaθ2 + bθ + c = 0. (3.2.4)
Writing√b2 − 4ac = r
√D with r ∈ Z, then
L = Q(D). (3.2.5)
If D < 0, then L is an imaginary quadratic number field.
January 23, 2014 65 Ali Nassar
3.3. The endomorphism ring of elliptic curves
3.3 The endomorphism ring of elliptic curves
Consider an elliptic curve (or a torus) Eτ = C/Λ, where Λ is a lattice Λ = (Z+ τZ)
and τ is the complex structure modulus. The endomorphisms of Eτ are given by
holomorphic maps F : Eτ → Eτ . The only such holomorphic maps fixing 0 are
induced by maps G : C → C of the form G(z) = λz where λ ∈ C is a constant
such that λΛ ⊆ Λ. Any elliptic curve will admit a set of trivial endomorphisms
corresponding to multiplication by integers λ ∈ Z, since multiplication by integers
just moves one from one lattice point to another. However, some special elliptic curves
could have non-trivial endomorphisms for particular values of τ . We want to find the
conditions on τ which give non-trivial endomorphisms.
Theorem 3.3.1. Let Eτ be an elliptic curve over C corresponding to the lattice Λ.Then
End(Eτ ) ∼= λ ∈ C : λΛ ⊂ Λ (3.3.1)
For λ ∈ End(Eτ ), then there exists j, k,m, n ∈ Z such that the action of λ on the
lattice Λ is represented on the generators as
λω1 = jω1 + kω2
λω2 = mω1 + nω2
(3.3.2)
here we recovered the lattice vectors ω1, ω2 to make the derivation clearer but we can
as well have worked with the basis 1, τ .
Writing (3.3.2) as (λ− j −k−m λ− n
)(ω1
ω2
)=
(0
0
)(3.3.3)
The existence of non-trivial solutions of the above equation implies that
λ2 − (j + n)λ+ (jn− km) = 0, (3.3.4)
January 23, 2014 66 Ali Nassar
3.3. The endomorphism ring of elliptic curves
that is, λ is an algebraic integer.
Now, we assume that λ is real, λ ∈ R. Using the fact that ω1 and ω2 are linearly
independent over R. Then the first equation in (3.3.2)
(λ− j)ω1 − kω2 = 0 (3.3.5)
implies that λ = j, that is λ ∈ Z which gives the set of trivial endomorphisms of any
elliptic curve.
If, on the other hand, λ is complex. Then λ ∈ C is an algebraic integer given by
λ =(j + n)±
√(j − n)2 + 4km
2(3.3.6)
Using this value of λ in (3.3.5) and after dividing by ω1 we get
τ =(n− j) +
√(j − n)2 + 4km
2k(3.3.7)
Define b = j − n, c = −m, and a = k and diving by the greatest common factor of
b = j − n, c = m, and a = k if necessary then we can write
τ =−b+
√b2 − 4ac
2a=−b+
√D
2a, (3.3.8)
where D = b2 − 4ac < 0. Hence
τ ∈ Q(D). (3.3.9)
From the above discussion we conclude that for generic τ , the endomorphism ring
End(Eτ ) is given by multiplication by integers. If, however, τ belongs to an imaginary
quadratic number field τ ∈ Q(D) then the elliptic curve has a bigger endomorphism
ring given by multiplication by numbers of the form (3.3.6). Elliptic curves (or tori)
for these special values of τ are said to have complex multiplication (or to be of CM
type).
January 23, 2014 67 Ali Nassar
3.4. Complex multiplication for higher-dimensional tori andCalabi-Yau manifolds
We end this section with what is called the first main theorem of complex mul-
tiplication. This theorem says that for τ ∈ Q(D), then j(τ) is an algebraic integer,
i.e.,
jh + a1jh−1 + · · ·+ ah = 0, ai ∈ Z, (3.3.10)
where h = h(D) is the class number of the field Q(D)1. We see that although j(τ) is
a non-trivial function of τ , for elliptic curves with complex multiplication j(τ) has a
very simple property.
3.4 Complex multiplication for higher-dimensional
tori and Calabi-Yau manifolds
The non-trivial endomorphisms of higher-dimensional tori can be found by a direct
generalization of (3.3.2). Let Z stand for the column vector made of the complex
coordinates zi, i = 1, · · · , n. The torus T n admits complex multiplication if it has
non-trivial endomorphisms [32]
Z −→ AZ. (3.4.1)
The condition on A can be derived in similar way as we did for the case of the 2-torus
A =M +NT
T A =M ′ +N ′T ,(3.4.2)
where M,N,M ′, N ′ are integer matrices. This gives the following second order equa-
tion for the period matrix
T NT + TM −N ′T −M ′ = 0. (3.4.3)
For Calabi-Yau manifolds, we first think of the period matrix in (3.1.44) as rep-
resenting some torus. The Calabi-Yau is said to have complex multiplication if the
1See the end of this chapter for the definition h = h(D)
January 23, 2014 68 Ali Nassar
3.5. Complex multiplication for higher-dimensional tori andCalabi-Yau manifolds
torus represented by (3.1.44) admits complex multiplication. This simply means that
T in (3.1.44) satisfies (3.4.3).
An example of a Calabi-Yau manifold which enjoys complex multiplication is the
Fermat quintic
P (z) =5∑
a=1
(za)5 = 0. (3.4.4)
The period matrix of the Fermat quintic is given by [75,76]
T =
(α− 1 α + α3
α + α3 −α4,
)(3.4.5)
where α5 = 1.
The matrix T satisfies (3.4.3) with
N =
(1 −10 1
), M =
(0 0
0 0
), N ′ =
(−1 0
−1 0
), M ′ =
(−1 0
−1 −1
).
(3.4.6)
The endomorphism matrix A is given by
A =
(α− 1 α + α3
1 + α + α3 −α4
)(3.4.7)
We note that α is an algebraic integer since it is a solution of a polynomial equation
with integer coefficients
α4 + α3 + α2 + α+ 1 = 0. (3.4.8)
As such, the elements of T and A belongs to
S = Q(α) (3.4.9)
which is the field of rational numbers extended by α.
In [32], Gukov and Vafa conjectured that a σ-model on a Calabi-Yau manifold
M gives rise to an RCFT if and only if M and its mirror W both have complex
multiplication over the same number field.
January 23, 2014 69 Ali Nassar
3.5. Quadratic Forms
3.5 Quadratic Forms
An integral quadratic form Q(a, b, c) in two variables x, y is defined as:
Q(a, b, c) : f(x, y) = ax2 + bxy + cy2 a, b, c ∈ Z. (3.5.1)
The quadratic form Q is called primitive if the greatest common divisor gcd(a, b, c) =
1. The discriminant of the quadratic form Q is D = b2 − 4ac.
We can associate to Q a 2× 2 matrix
M(Q) =
(2a b
b 2c
). (3.5.2)
The discriminant of Q is now given by D = − det(M).
The matrix representation allows us to define an equivalence relation. Two forms
Q and Q′ of the same discriminant are equivalent Q ∼ Q′ if there exists S ∈ SL(2,Z)
such that
M(Q′) = STM(Q)S. (3.5.3)
Since S = ±I acts trivially, then only the group PSL(2,Z) acts effectively. The above
action of SL(2,Z) leaves D unchanged. Two quadratic forms are said to be properly
equivalent if they are in the same SL(2,Z) orbit. Now we can talk of equivalence
classes of quadratic forms under the action of SL(2,Z). We simply take all the
matrices which are in the same orbit of Q under SL(2,Z) as one equivalence class.
The set of properly SL(2,Z)-equivalent classes is defined as [33]
Cl(D) =Q(a, b, c) : primitive quadrartic form
∣∣D = b2−4ac < 0, a > 0/ ∼ SL(2,Z).
(3.5.4)
The set Cl(D) is finite and its cardinality h(D) = |Cl(D)| is the class number of
January 23, 2014 70 Ali Nassar
3.5. Quadratic Forms
Q(D). The set of classes in Cl(D) will be written as
Cl(D) = C1, · · · , Ch(D). (3.5.5)
We can repeat the same construction to the group GL(2,Z) and define the set of
improperly GL(2,Z)-equivalent classes
Cl(D) =Q(a, b, c) : primitive quadrartic form
∣∣D = b2−4ac < 0, a > 0/ ∼ GL(2,Z).
(3.5.6)
Now, let us study the quadratic form
Q = f(τ, 1) = aτ 2 + bτ + c. (3.5.7)
We only consider the values of τ in the upper-half plane
H+ = τ ∈ C | ℑ(τ) > 0, (3.5.8)
that is
τ =−b+
√D
2a(3.5.9)
We can associate with each quadratic form Q(a, b, c) a complex number τQ
τQ(a,b,c) =−b+
√D
2a. (3.5.10)
We have the map
Q(a, b, c) 7→ τQ(a,b,c) ∈ H+. (3.5.11)
The SL(2,Z) action on Q induces a fractional linear transformation on τQ. One can
parametrize the equivalence classes of quadratic forms by the points inH+/PSL(2,Z).
The SL(2,Z) action on quadratic forms is compatible with the PSL(2,Z) on H+. The
PSL(2,Z) orbits of τQ(a,b,c) ∈ H+ depends on the class C = [Q(a, b, c)] ∈ Cl(D).
January 23, 2014 71 Ali Nassar
3.5. Quadratic Forms
Quadratic forms and their equivalence classes are related to rank-2 lattices. We
consider an even integer lattice Γ with basis e1 and e2 and an intersection form K.
The intersection form K can be written as(2a b
b 2a
):=
(⟨e1|e1⟩ ⟨e1|e2⟩⟨e2|e1⟩ ⟨e2|e2⟩
), (3.5.12)
where a, b ∈ Z. Since Γ is positive definite, then 4ac− b2 = −D > 0 and a > 0. We
say that Γ is primitive if gcd(a, b, c) = 1 which is equivalent to a primitive quadratic
form.
January 23, 2014 72 Ali Nassar
Chapter 4
Strings on CM tori
We have seen that CFTs based on CM tori are rational. Hence, one should expect that
there is an extended chiral algebra underlying those RCFTs. A simple demonstration
of this is the Moore-Seiberg algebra for the c = 1 boson on a rational circle (2.2.2).
One can try to write down a Moore-Seiberg algebra for the rational c = 2 theory as
for the c = 1 case. However this algebra is already complicated for the c = 1 case
(see (2.2.2)). In this thesis, we propose using a simpler, more compact chiral algebra
to describe the RCFTs which describe strings on CM tori. These are the matrix-level
affine Kac-Moody algebras studied by Gannon [34].
In this chapter, we establish the correspondence between strings on a CM tori and
the U2,K algebra. The algebra U2,K is characterized by the left and right matrix-valued
levels KL and KR. We will use the Gauss product to construct KL and KR from the
complex structure and the Kahler structure parameters τ and ρ which characterize
strings on tori. For CM tori, τ and ρ are takes values in a quadratic number field
τ, ρ ∈ Q(D). This enables one to represent τ and ρ by binary quadratic forms where
a Gauss product is defined.
The special values of the moduli indicate RCFTs where the infinite number of
73
4.1. CHAPTER 4. STRINGS ON CM TORI
fields is organized into a finite set that is primary with respect to an extended chiral
algebra [48]. In the Tm case, the generic boson algebra U(1)m = U(1)k1×· · ·×U(1)kmis extended by vertex operators defined by vectors of the lattice Λm describing Tm ∼=
Rm/Λm. The extended algebra is well understood; it was written explicitly for the
m = 1 case in [48], for example.
On the other hand, a different algebra has also been of interest. The Abelian
algebra Um,K with an m × m matrix-valued level K was studied in [34], where its
modular invariant partition functions were classified. This matrix-level algebra gener-
alizes U(1)m = U(1)k1×· · ·×U(1)km , which is recovered for a diagonal matrixK. Here
we consider the more general case, allowing K to be a non-negative integer-valued
matrix.
This relation has already proven useful in the following way. As was noted in [34],
the moduli space of the RCFTs based on the Um,K algebra is given in terms of
the moduli space of rational points on the Grassmannian Gd,d(R). This is similar
to the Narain moduli space of compactifications of strings on tori. We show that
the characterization of the Um,K partition functions in terms of rational points on a
Grassmannian [34] is equivalent to specifying CM tori inside the Narain moduli space.
This is another way to show that the set of RCFTs is dense in the Narain moduli
space since the set of rational points is dense in the Grassmannian Gd,d(R). We hope
that the relation between matrix-level RCFTs and strings on tori will also be helpful
in other ways. The results in this chapter are published in [77].
January 23, 2014 74 Ali Nassar
4.1. Matrix-level affine algebras
4.1 Matrix-level affine algebras
In this section, we study affine Abelian chiral algebras with matrix-valued level in
more details. To provide evidence that they are the chiral algebras of a consistent
class of RCFTs, we describe their modular invariant partition functions [34]. In
Section 4.2 we relate them to σ-models on CM tori.
Let ΓK be a Euclidean, even, integral lattice of rank r
ΓK = Ze1 + · · ·+ Zer, ⟨ei|ej⟩ = Kij ∈ Z, (4.1.1)
where positive definite integer-valued symmetric intersection matrix Kij has a deter-
minant −D > 0. In the following we will refer to a lattice ΓK and its intersection form
K interchangeably with the hope that this will not cause any confusion. The basis
ei of ΓK is defined up to GL(r,Z) transformations which preserve the determinant
of K
ei → Ajiej, A ∈ GL(r,Z) , detA = ±1 . (4.1.2)
There are r linearly independent U(1) currents
Ji(z) =∑n∈Z
Jni
zn+1. (4.1.3)
They have the OPE
Ji(z)Jj(w) ∼Kij
(z − w)2. (4.1.4)
This is the Abelian version of (2.8.2). In terms of the modes we have[Jmi , J
nj
]= mKijδm+n,0. (4.1.5)
The zero modes of the currents can be used to grade the Hilbert space into different
superselection sectors with different charges q ∈ ΓK
H = ⊕qHq, (4.1.6)
January 23, 2014 75 Ali Nassar
4.1. Matrix-level affine algebras
where
Hq =|q⟩ | J0
i |q⟩ = qi|q⟩ (4.1.7)
The energy momentum tensor is given by the Sugawara construction
T (z) =1
2Kij : Ji(z)Jj(z) :, (4.1.8)
where Kij = K−1ij is the intersection form of the dual lattice.
From the OPE of T (z) with itself we can read off the central charge
c = KijKij = r. (4.1.9)
The Virasoro generators are given by
Ln =1
2Kij
∞∑m=−∞
: Jm+ni J−m
j : . (4.1.10)
The Virasoro ground state in the sector Hq is defined by
L0|q⟩ =1
2q2|q⟩, Ln|q⟩ = 0, (4.1.11)
Now we specialize to rank two lattices ΓK with basis e1, e2. Since the lattice is
even-integer lattice, then its intersection form can be written as
Kij = ⟨ei|ej⟩ =
(2a b
b 2c
), a, b, c ∈ Z. (4.1.12)
We assume that gcd(a, b, c) = 1 which corresponds to primitive quadratic forms.
The GL(2,Z) transformation on the basis of ΓK gives an equivalent lattice. The
set of equivalence classes of primitive, even lattices is defined as
Lp(D) :=ΓK : D = − detK
/GL(2,Z). (4.1.13)
We will consider different matrix levels KL for the holomorphic and KR for the
anti-holomorphic sectors which give rise to heterotic theories. The set of standard
January 23, 2014 76 Ali Nassar
4.1. Matrix-level affine algebras
representations of the affine algebra Um,K are labelled by a ∈ PKL+ = Γ∗
KL/ΓKL
and
b ∈ PKR+ = Γ∗
KR/ΓKR
where |KL| = |KR| and |K| is the determinant of K [34].
The RCFT data are given by (ΓKL,ΓKR
, χΓKLa , χΓKR
a ) where χΓKRa are the
characters which are proportional to the theta functions of the lattice
χΓKLa (q) =
θΓKLa (q)
η(q)2=
1
η(q)2
∑v∈ΓKL
q12(a+v)2 , (4.1.14)
where η(q) is the Dedekind eta function.
We can compute the modular S matrix from transformation of the characters
(χΓKLa = θ
ΓKLa /η2) under τ → −1/τ
χΓKLa
(−1τ
)=θΓKLa (−1/τ)η(−1/τ)2
. (4.1.15)
Using the transformation properties of the theta functions we find
Sab =i√De−2iπ(a·b). (4.1.16)
Now we can find the fusion rules using the Verlinde formula
N cab =
∑d
SadSbdS−1cd
S0d
(4.1.17)
Since the gluing map (see the discussion after (4.2.14)) satisfies (a, b) = (φ(a), φ(b))
then we see that
N cab = N
φ(c)φ(a)φ(b), (4.1.18)
i.e., φ is a symmetry of the fusion rules and hence can be used to build new modular
invariants. The quantum dimension of a primary field labelled by a is
Sa0
S00
= 1, (4.1.19)
January 23, 2014 77 Ali Nassar
4.1. Matrix-level affine algebras
i.e., all the primaries are simple currents. This simplifies the construction of all
modular invariants of the algebra as was shown in [34].
The spectrum is encoded in the (genus-1) partition function
ZΓKL,ΓKR =
∑a∈PKL
+ , b∈PKR+
Ma,b χΓKLa χ
ΓKRb , (4.1.20)
where the matrix Ma,b is constrained to satisfy Ma,b ∈ Z≥ and M0,0 = 1.
Modular invariance dictates that
SM =MS, TM =MT, (4.1.21)
where the matrices S and T are unitary and symmetric and T is diagonal
Sab =1√|KL|
exp[−2πi(a · b)]. (4.1.22)
In [34], all modular invariants of the algebra Um,K were constructed in terms of
even self-dual lattices Γ that contain ΓKLand ΓKR
. Here we are only concerned with
physical modular invariants, i.e., those which satisfy (4.1.21)1. In terms of the matrix
K, full modular invariance means Kij ∈ 2Z. This, together with the symmetry of K,
translates to an even integer lattice ΓK .
The argument in [34] goes as follows: consider the heterotic partition function
(4.1.20). Using SM =MS and the unitarity of the S matrix we find
Ma,b =∑c,d
Sa,cMc,dS∗d,b
=1
|KL|∑c,d
exp[2πi(b · d− a · c)]Mc,d.(4.1.23)
1In [34], the second constraint in (4.1.21) is relaxed which leads to a much bigger set of modularinvaraints which include weak modular invariants. But since we are looking for a geometric interpre-tation in terms of a torus target we will require full modular invariance. The σ-model with a torustarget is modular invariant by construction.
January 23, 2014 78 Ali Nassar
4.1. Matrix-level affine algebras
From this equation we derive the relation
Ma,b
M0,0
=
∑c,d zab,cdMc,d∑
c,dMc,d
, (4.1.24)
where we defined
zab,cd = exp[2πi(b · d− a · c)], |zab,cd| = 1. (4.1.25)
Using the triangle inequality, then 4.1.24 implies that |Ma,b| ≤ |M0,0| = 1 with the
equality iff
Mc,d = 0 =⇒ b · d = a · c (mod 1) (4.1.26)
for all c ∈ PKL+ , d ∈ PKR
+ . The above equation means that any two labels which
appear in the character decomposition of the partition function must be related by
an isometry of the discriminant groups b = φ(a) and c = φ(d) andMa,b = δb,φ(a). The
partition function now can be written as
ZΓKL,ΓKR =
∑a∈PKL
+
χΓKLa χ
ΓKR
φ(a) , (4.1.27)
which is an example of an automorphism modular invariant. We only consider au-
tomorphism modular invariant partition functions of the algebra Um,K . If Um,K is
the maximally extended chiral algebra of strings on CM-tori then the automorphism
modular invariants will exhaust the set of all modular invariants. However, if Um,K is
not the maximally extended chiral algebra then there will be other non-automorphism
modular invariants.
Define the set
Ω =∪
a∈PKL+ , b∈PKR
+
(a⊕ ib) +(ΓKL⊕ iΓKR
)(4.1.28)
January 23, 2014 79 Ali Nassar
4.2. Strings on a CM torus
which is an even self dual lattice. The matrix Ma,b in (4.1.20) satisfies
Ma,b =
1 if (a, ib) ∈ Ω
0 otherwise.(4.1.29)
Hence, the modular invariant partition functions of the Um,K current algebra are in
1-to-1 correspondence with the even self-dual lattices Γ which contain (ΓKL; ΓKR
).
4.2 Strings on a CM torus
We consider the compactification of string theory on an elliptic curve Eτ which has
complex multiplication (or a CM torus). The σ-model on Eτ is specified by the
parameters τ and ρ. The compactification of strings on Eτ is characterized by a
momentum-winding Narain lattice, an even self-dual lattice Γ(τ, ρ) of rank 4, where
the parameters τ and ρ live in the upper-half plane H+ subject to a group of discrete
symmetries Ξ. The Narain moduli space of this compactification is [32]
M =H+ ×H+
Ξ, (4.2.1)
where Ξ is the group of discrete symmetries of Eτ
Ξ = PSL(2,Z)τ × PSL(2,Z)ρ × Z2 × Z2 × Z2. (4.2.2)
The first Z2 is a mirror symmetry which exchanges τ and ρ, i.e., Z2 : (τ, ρ) 7→ (ρ, τ).
The second Z2 is a space-time parity transformation Z2 : (τ, ρ) 7→ (−τ ,−ρ), where
the bar denotes complex conjugation. The last Z2 is a world-sheet orientation reversal
Z2 : (τ, ρ) 7→ (τ,−ρ)2.2The three Z2 symmetries are not really independent since the second Z2 is equivalent to applying
the third Z2 then the first Z2 then the third Z2 then the first Z2 again.
January 23, 2014 80 Ali Nassar
4.2. Strings on a CM torus
For generic values of τ and ρ, the conformal field theory is not rational and has an
infinite number of primary fields. For special values of τ and ρ the infinite number of
primary-field representations reorganize into a finite set of representations of a bigger
chiral algebra and the theory becomes rational. It was shown in [32] that CFTs based
on Eτ are rational iff Eτ and its mirror are both of CM type, that is, τ, ρ ∈ Q(D)
which implies that ρ, like τ , satisfies a quadratic equation with integer coefficients
a′ρ2 + b′ρ+ c′ = 0, ρ =−b′ +
√D
2a′, D′ = b′2 − 4a′c′ < 0. (4.2.3)
An example of an RCFT which enjoys this property results when τ = ρ = e2πi/3.
In this case the chiral vertex operator algebra is isomorphic to that of SU(3) WZW
model at level 1 [78]. Clearly, both τ and ρ satisfy
τ 2 + τ + 1 = 0. (4.2.4)
In [32], the diagonal case was studied in detail where it was shown that the
condition for a diagonal modular invariant is
τ = faρ, (4.2.5)
where f ∈ Z, τ, ρ ∈ Q(D), and a is the coefficient of τ 2 in
aτ 2 + bτ + c = 0. (4.2.6)
The generalization to non-diagonal modular invariants was given in [33].
We can associate a quadratic form with the complex numbers τ and ρ. Write
Q(a, b, c) =
(2a b
b 2c
), τQ(a,b,c) =
−b+√D
2a(4.2.7)
January 23, 2014 81 Ali Nassar
4.2. Strings on a CM torus
where D = b2 − 4ac = − detQ denotes the discriminant of the quadratic form Q. Q
is called primitive if gcd(a, b, c) = 1. The discriminant of the quadratic form Q is
invariant under SL(2,Z):
Q −→ StQS, S ∈ SL(2,Z). (4.2.8)
Now we can consider equivalence classes of quadratic forms under the action of
SL(2,Z) (or more precisely PSL(2,Z), since S = ±I acts trivially). The set of
equivalence classes is denoted by
Cl(D) =Q(a, b, c)
∣∣D = b2 − 4ac < 0, a > 0/ ∼ SL(2,Z). (4.2.9)
It is known that Cl(D) is a finite set and we will denote the number of its elements
by h(D) (see [33] and references therein)
Cl(D) =C1, . . . , Ch(D)
. (4.2.10)
Since D < 0, the complex number τQ lies in the upper-half plane H+. The SL(2,Z)
action on Q will induce a fractional linear transformation on τQ. The PSL(2,Z)
action on quadratic forms is compatible with the PSL(2,Z) on H+. The PSL(2,Z)
orbits of τQ(a,b,c) ∈ H+ depend on the class C = [Q(a, b, c)] ∈ Cl(D). Using the above
mapping, then we can label the classes in Cl(D) by points [τQ] ∈ F = H+/PSL(2,Z).
Since τ is the complex structure of a torus then fractional linear transformations on
τ gives an equivalent torus. The classes [τQ] will give inequivalent tori in the Narain
moduli space.
The same goes for ρ where we can also define equivalence classes of quadratic
forms parametrized by points [ρQ′ ] ∈ F = H+/PSL(2,Z). The equivalence class of
Narain lattices corresponding to an RCFT will be denoted by Γ(τC, ρC′), where C and
C ′ are the equivalence classes corresponding to τ and ρ.
January 23, 2014 82 Ali Nassar
4.2. Strings on a CM torus
Similarly, the group GL(2,Z) acts on quadratic forms by the same formula as
(4.2.8). The set of improper equivalence classes under GL(2,Z) is defined by [33]
Cl(D) =Q(a, b, c)
∣∣D = b2 − 4ac < 0, a > 0/ ∼ GL(2,Z). (4.2.11)
and we have a surjection q
q : Cl(D)→ Cl(D), C → C, (4.2.12)
where q−1(C) has either one or two classes.
One can associate with the lattice ΓK a primitive quadratic form given by the
intersection form K. Therefore, the equivalence classes of primitive lattices [ΓK ] are
in 1-to-1 correspondence with the equivalence classes of primitive, quadratic forms
[Q(a, b, c)]. Hence, we can identify the set Cl(D) with the set Lp(D) in (4.1.13).
We define the following projections of Γ(τC, ρC′)
ΠL := Γ(τC, ρC′) ∩ R2,0, ΠR := Γ(τC, ρC′) ∩ R0,2, (4.2.13)
which correspond to the equivalence classes of the left and right momentum lattices,
characterized by the vanishing of the right moving and left moving momenta, respec-
tively. 3
The modular invariant partition functions studied in [33] take the form:
ZΠL,ΠR,φ(q, q) =1
|η(q)|4∑
a∈Π∗L/ΠL
θΠLa (q), θΠR
φ(a)(q)
=∑
a∈Π∗L/ΠL
χΠLa (q), χΠR
φ(a)(q)
(4.2.14)
where φ is a gluing map between the discriminant groups Π∗L/ΠL and Π∗
R/ΠR. It
satisfies (φ(a), φ(b)) = (a, b), where a, b ∈ Π∗L/ΠL and (·, ·) is the rational bilinear
form on Π∗L/ΠL which is induced from the bilinear form on ΠL.
3More precisely it corresponds to an equivalence class of lattices.
January 23, 2014 83 Ali Nassar
4.2. Strings on a CM torus
The characters which enter the partition function above are given in terms of the
theta function of the lattice ΠL:
χΠLa (q) =
θΠLa (q)
η(q)2=
1
η(q)2
∑v∈ΠL
q12(a+v)2 (4.2.15)
They are identical to the characters in (4.1.14).
4.2.1 The Gauss product
There is a binary operation which turns the set Cl(D) of (4.2.9) into an Abelian
group: the Gauss product (see [33], e.g.) takes two equivalence classes of quadratic
forms of the same discriminant and produces a third with that discriminant.
Let C = [Q1(a1, b1, c1)] and C ′ = [Q2(a2, b2, c2)] be two such equivalence classes.
We will restrict ourselves to primitive forms. We say that two quadratic forms
Q1(a1, b1, c1) ∈ C and Q2(a2, b2, c2) ∈ C ′ are concordant if a1a2 = 0, gcd(a1, a2) = 1
and b1 = b2. Then the Gauss product of C ⋆ C ′ is defined as
[Q1
(a1, b, c1
)]⋆[Q2
(a2, b, c2
)]=[Q3
(a3, b3, c3
)], (4.2.16)
where a3 = a1a2, b3 = b, and c3 = b2−D4a1a2
. It is important to mention that any
pair of quadratic forms of the same discriminant can be SL(2,Z)-transformed to a
concordant pair.
The unit 1D of Cl(D) with respect to the product ⋆ is represented by
1D =
[1, 0,−D
4]; if D ≡ 0 mod 4
[1, 0, 1−D4
]; if D ≡ 1 mod 4.
(4.2.17)
The quadratic form Q3(a3, b3, c3) corresponds to a lattice with intersection form
Q3
(a3, b3, c3
)≡
(2a3 b3
b3 2c3
). (4.2.18)
January 23, 2014 84 Ali Nassar
4.2. Strings on a CM torus
which gives an even integer lattice, an important fact which we will use when we
construct the matrix levels KL and KR for the Um,K algebras.
The Gauss product is used to construct the intersection form of the lattices ΠL
and ΠR in terms of the equivalence classes C and C ′, corresponding to τQ(a1,b1,c1) and
ρQ(a2,b2,c2), respectively [33]:
ΠL = q(C ⋆ C ′−1), ΠR = q(C ⋆ C ′)(−1). (4.2.19)
Here q is the natural map Cl(D) → Cl(D) and
q(C ⋆ C ′)(−1) means we multiply the quadratic form q(C ⋆ C ′) by −1.
As was shown in [33], to prove the above result one first constructs the Z-basis
for ΠR of Γ(τC, ρC′) in terms of the equivalence classes of C and C ′ and then compares
the resulting quadratic form of ΠR with Q3(a3, b3, c3) and similarly for ΠL.
In the special case when τ ∝ ρ, we have ΠL = ΠR = Π and we get a diagonal
modular invariant for φ = I
ZΠ,Π,I(q, q) =1
|η(q)|4∑
a∈Π∗/Π
θΠa (q)θΠa (q). (4.2.20)
Now we have geometric data represented by the rational Narain lattice Γ(τC, ρC′)
which depends on the σ-model parameters τ and ρ (both τ and ρ are attached to
an equivalence class of quadratic forms) and algebraic, RCFT data (ΓKL,ΓKR
, χa).
The Gauss product can be used to relate them in the following way. First we will
look at KL and KR as quadratic forms and hence we can talk about their respective
equivalence classes under the SL(2,Z) action, as we did with Q. Now, a rational
point in the Narain moduli space specified by the special values τC and ρC′ defines
two equivalence classes of quadratic forms C and C ′. The mapping between the two
January 23, 2014 85 Ali Nassar
4.2. Strings on a CM torus
sets of data is
KL = q(C ⋆ C ′−1
), KR = q
(C ⋆ C ′
)(−1). (4.2.21)
We will need to apply a symmetrisation map (half the sum of the matrix and its
transpose) to KL and KR if they are not symmetric or SL(2,Z)-equivalent to their
symmetric forms.
The mapping (4.2.21) shows that the algebras Um,K can be given a geometric
significance, by relating them to σ-models on CM tori. To justify this, we notice
that the characters of the RCFTs in (4.2.15) which are proportional to the theta
functions of the momentum-winding lattice are the same as the characters (4.1.14) of
the algebra U2,K . Both sets of characters are based on lattices which are constructed
from the same set of geometric data using the Gauss product. Also, the modular
invariant partition functions (4.2.14) are a subset of the modular invariant partition
functions (4.1.20) of the algebra U2,K for which
Ma,b = δb,φ(a). (4.2.22)
We note that the above mapping is not 1-to-1. We can’t start from KL and KR
and construct unique equivalence classes for τ and ρ. The matrices KL and KR are
representatives of equivalence classes in the set Cl(D) which have a finite number of
elements. The algebraic description of the Um,K algebras depends only on KL and
KR and as such is the same for all members of the set Cl(D). On the other hand
there is a geometric description for each member of the set Cl(D). We conclude that
the same Um,K algebra have many geometric avatars. This is similar to the case of
a rational boson on a circle of radius square R2 = p/q which is described by a level
k = pq U(1)k algebra. On the other hand, starting from the algebra U(1)k, there
January 23, 2014 86 Ali Nassar
4.3. Wen topological order
are many candidate rational boson theories, one for each factorization of k into two
coprime integers k = pq.
4.3 Wen topological order
As a byproduct of the geometric connection of the Um,K algebras we can relate the
Wen topological order [70,71,79] to the number of D0 branes in the models discussed
in this chapter. It was shown in [32] that the number of D0 branes on T 2 is equal to
the dimension of the chiral ring which is given by Γ∗L/ΓL.
Topological order is a new kind of order which arise in the topological phases
of many systems, e.g., fractional quantum Hall systems. It is not related to any
symmetry breaking, like the usual ordered-disordered transitions described by the
Landau theory. There is no local order parameter which takes a non-zero vacuum
expectation value in the broken phase. A new set of quantum numbers needs to be
defined which characterizes topological order.
Wen topological order is the degeneracy of the quantum Hall ground state on the
compact surface Σg of genus g and is one of the quantum numbers which characterizes
the topological order. The low energy effective field theory of the quantum Hall
ground state is the Chern-Simons theory on M = Σ × R. The degeneracy of the
ground state is counted by the dimension of the Hilbert space of topologically non-
trivial gauge fields on M [48, 80, 81]. This dimension can be readily computed from
the RCFT fusion rules and the S modular transformation. The general formula for
the dimension of the space of conformal blocks on a genus g surface with n punctures
January 23, 2014 87 Ali Nassar
4.4. Rational points on Grassmannians and CM tori
was found in [82,83]
dimH(Σg; (P1, i1), · · · , (Pn, in)) =∑
a∈Π∗L/ΠL
(1
Sa0
)2(g−1)Ski1
Sa0
· · ·Sain
Sa0
. (4.3.1)
For the torus g = 1 with no punctures and using the fact that all the primaries are
simple currents we get
dimH(Σ1) =∑
a∈Π∗L/ΠL
1 = D. (4.3.2)
which is simply the dimension of the chiral ring identified with the number of D0
branes in [32].
4.4 Rational points on Grassmannians and CM
tori
In this section we study the rationality conditions of a Narain lattice in more detail.
We formulate the rationality in terms of rational points on a Grassmannian and we
show that these points are equivalent to tori of CM type. Our argument will be based
on the results in [30].
We consider a generic Narain lattice Γ(τ, ρ) = (PL;PR) of the σ-model on T d/Λ
with a B-field
Γ(τ, ρ) =(PL;PR
):=
1√2
(µ− Bλ+ λ;µ− Bλ− λ
)| (µ, λ) ∈ Λ∗ ⊕ Λ
, (4.4.1)
where Λ∗ and Λ are the momentum and winding lattices, respectively and B = ΛT BΛ.
The holomorphic and anti-holomorphic vertex operators are characterized by PR =
0 and PL = 0 and they are parametrized by the values of their charges in ΠL = (PL; 0)
January 23, 2014 88 Ali Nassar
4.4. Rational points on Grassmannians and CM tori
and ΠR = (0;PR)
ΠL =(PL; 0
):=
1√2
(µ− Bλ+ λ; 0
)|(µ, λ) ∈ Λ∗ ⊕ Λ
ΠR =
(0;PR
):=
1√2
(0;µ− Bλ− λ
)|(µ, λ) ∈ Λ∗ ⊕ Λ
.
(4.4.2)
We also define the following projections of the lattice Γ(τ, ρ):
ΠL = (PL; ∗) :=1√2
(µ− Bλ+ λ; ∗
)|(µ, λ) ∈ Λ∗ ⊕ Λ
ΠR = (∗;PR) :=
1√2
(∗;µ− Bλ− λ
)|(µ, λ) ∈
(4.4.3)
where the ∗ means we forget about the corresponding component of P .
Note that
ΠL ⊆ ΠL, ΠR ⊆ ΠR. (4.4.4)
Since the lattice Γ(τ, ρ) is even, self-dual and integral then its straightforward to show
that
Π∗L∼= ΠL, Π∗
R∼= ΠR, (4.4.5)
i.e., ΠL is the dual of ΠL and the same for ΠR.
Rationality can be expressed in terms of the rank of ΠL and ΠR . The Narain
lattice Γ(τ, ρ) is rational if and only if [30]
rank(ΠL
)= rank
(ΠR
)= d. (4.4.6)
RCFTs are characterized by the appearance of extra holomorphic vertex operators
which extend the chiral algebra. For a generic Narain lattice Γ(τ, ρ), the only holo-
morphic vertex operator is the one corresponding to the vacuum PL = PR = 0. Since
any field is mutually local 4 to the vacuum, then the set of allowed representations
4Two fields are mutually local if their OPE doesn’t contain branch cuts. In other words the product of their charges is integer.
January 23, 2014 89 Ali Nassar
4.4. Rational points on Grassmannians and CM tori
V = V [P ], P ∈ Γ(τ, ρ) is infinite. However extra holomorphic vertex operators
W [P ] appear for p ∈ ΠL. The requirement of locality with respect to W [P ] restricts
the set of allowed irreducible representations to be a ∈ Π∗L/ΠL, where Π
∗L is the lattice
dual to ΠL so it contains charges which have integer product with ΠL. The condition
for rationality translates to the requirement that ΠL be a finite index sublattice of Π∗L
so that the set Π∗L/ΠL has a finite cardinality. This happens when ΠL have a finite
rank which is the condition in (4.4.6).
The Narain moduli space of conformal field theories is isomorphic to the moduli
space of even, self-dual lattices with signature (d, d)
Md = O(Γd,d)\O(d, d)/(O(d)×O(d)), (4.4.7)
where Γd,d denotes the standard even self-dual lattice of signature (d, d) and O(Γd,d)
its automorphism group. In the special case of d = 2 this gives the moduli space in
(4.2.1). The above moduli space is also the Grassmannian of space-like d-planes in
Rd,d. The modular invariant partition functions in [34] are classified using rational
points on the Grassmannian (4.4.7).
Note that
PL ∈ W = ΠL ⊗ R, PR ∈ W⊥ = ΠR ⊗ R, (4.4.8)
where W is a space-like d-plane which correspond to a point on the Grassmannian
(4.4.7) and W⊥ is its orthogonal complement in Rd,d.
The sets ΠL and ΠR in general are not lattices, since the set of vectors which
span ΠL and ΠR will not remain linearly independent over Z when restricted to W
and W⊥. However, if W is a rational point on the Grassmannian (4.4.7) then ΠL
and ΠR become lattices of rank d. A rational point on the Grassmannian (4.4.7) is a
January 23, 2014 90 Ali Nassar
4.4. Rational points on Grassmannians and CM tori
subspace W with basis fm which can be written over Q in terms of the preferred
orthonormal basis ei of Rd,d
fm = Qmiei, Qmi ∈ Q. (4.4.9)
The above equation implies that the basis of W are rational vectors
⟨fm|fn⟩ ∈ Q. (4.4.10)
Since any group generated by rational vectors is a lattice, i.e., it can be generated by
linearly independent vectors over Z. Then ΠL which is the Z-span of fm is a lattice
of rank d = dim(W ) and the same for ΠR. It was shown in [30] that for d = 2
rank(ΠL
)= rank
(ΠR
)= 2←→ τ, ρ ∈ Q(D). (4.4.11)
which in our case implies that rational points on the Grassmannian (4.4.7) correspond
to CM tori. This is another way to see that RCFTs constitute a dense subset of the
set of CFTs, since the set of rational points on a Grassmannian is dense. This is
consistent with the finding in [32] that the values τ, ρ ∈ Q(D) corresponding to
RCFTs are dense in in the Narain moduli space.
January 23, 2014 91 Ali Nassar
Chapter 5
Conclusion
In summary, we studied the RCFTs based on the matrix-level algebra U2,K and we
related them to strings on CM tori (corresponding to τ, ρ ∈ Q(D)) inside the Narain
moduli space. The characters and modular-invariant partition functions were shown
to be identical in the 2 types of theories.1 Furthermore, the map between them was
constructed explicitly: the Gauss product was used to write the matrix levels KL and
KR in terms of the geometric data represented by τ and ρ.
The connection was shown to be useful in one way. By formulating the problem
in terms of rational points on a Grassmannian we showed that the set of RCFTs is
a dense subset in the Narain moduli space. This agrees with the observation in [32]
that the values of τ, ρ ∈ Q(D) which produce RCFTs are dense in the Narain moduli
space. We anticipate that the relation between matrix-level algebras and RCFTs
for strings on tori will prove useful in other ways. Using the relation between U2,K
and strings on CM tori, Wen topological order for fractional quantum hall states was
related to the number of D0 branes which are allowed on the CM torus.
The Gukov-Vafa criterion characterizes the rationality of the CFT in terms of the
1For characters, compare (4.2.15) to (4.1.14); for partition functions, (4.2.14) to (4.1.20,4.1.29).
92
5.0. CHAPTER 5. CONCLUSION
geometry of the target space. In the Gannon classification, rationality is derived from
the algebraic properties of the chiral algebra without reference to any target space
interpretation. We showed that the geometric and the algebraic characterization of
rationality are related and that the density property of RCFTs can be seen in both
pictures.
We also proposed a generalization of Um,K to the non-Abelian case. This would
be related to the Chern-Simons theories on a three-manifold N3, based on a gauge
group G with the level being an integer valued matrix
S = KAB
∫M4
FA ∧ FB, (5.0.1)
where M4 is the four manifold for which N3 is a boundary. The independence of the
partition function from the extension from N3 to M4 forces KAB to be an integer-
valued matrix.
It would be interesting to see if the results described here might be generalized
to higher dimensions m > 2, or to non-Abelian theories. The m = 2 case is special
because only in this case one has a Gauss product which turns the set of equivalence
classes of quadratic forms (or rank-2 lattices) into an Abelian group. For higher
dimension tori which can be written as a product of elliptic curves
Tm = Eτ1 × · · · × Eτm/2, (5.0.2)
one can use the Gauss product for each individual factor to construct the algebra
Um,K . But a generic Tm will not factorize into a product of elliptic curves and the
Gauss product can’t be used directly.
Most exciting, perhaps, might be an explanation of the mysterious connection
between bosons on a CM torus and matrix level. Does the chiral algebra of the
January 23, 2014 93 Ali Nassar
5.0. CHAPTER 5. CONCLUSION
former, extended as it is by vertex operators, have any more direct relation to the
(simpler) Abelian matrix-level algebras, e.g.?
One could try to understand the Gannon classification [34] from a 3D perspective
using the Chern-Simons dual theory. It would be interesting to derive the rationality
condition
aτ 2 + bτ + c = 0 (5.0.3)
from the Chern-Simons theory. The condition (5.0.3) which picks rational confor-
mal field theories doesn’t seem to be related to any physical property of the 2D
CFT. Hopefully the 3D Chern-Simons dual would give a more physical explanation
of (5.0.3). Our guess is that the condition (5.0.3) should arise as the minimum of a
potential V (τ) over the complex structure moduli space parametrized by τ . If this
the case, then we will learn that RCFTs are important in string compactification not
only because of their simplicity (or their density in the moduli space of all CFTs but
because they minimize some sort of energy on the space of CFTs.
Most importantly from the point of view of string theory is further investigation
of the Gukov-Vafa conjecture. One could try to find other Gepner points in the
moduli space of Calabi-Yau manifolds and check if these Calabi-Yau manifolds admit
complex multiplication. The conjecture is obviously valid for Calabi-Yau manifolds
which can be described as orbifolds of complex multiplication tori. It was also shown
in [32] that the conjecture is also valid for Calabi-Yau manifolds which correspond to
the Fermat point in the complex structure moduli space.
The connection with FQHE can be pursued further and the results in [77] can be
applied to the characterization of topological order in the fractional quantum Hall
states [70, 71, 79]. As we have seen in Section 2.8, the multi-layer FQH states are
January 23, 2014 94 Ali Nassar
5.0. CHAPTER 5. CONCLUSION
described by Laughlin wave functions which are labeled by a matrix K
ΨK(z) =∏a,b,i,j
(zai − zbj)Kab/2 exp(−∑|zai|2
). (5.0.4)
They are in 1-1 correspondence with SL(m,Z) equivalence classes of K matrices. For
m = 2, i.e., for a double layer FQH states, we can use the Gauss product to define a
product on the set of Laughlin wave functions
ΨK1(z)⊙ΨK2(z) = ΨK3(z), (5.0.5)
where K3 is given by the Gauss product of K1 and K2
K3 = K1 ⋆ K2. (5.0.6)
The set of Laughlin wave functions together with the product ⊙ now forms an Abelian
group G. The group property of ⊙ follows directly from the group property of ⋆. As
far as we know, this kind of group structure on the set of Laughlin wave functions
didn’t appear before in the study of FQHE. The dimension of G is equal to the number
of equivalence classes of K with determinant D = det(K) which is nothing but the
class number h(D). As we have seen in (4.3.2), D gives the degeneracy of the ground
state of the FQH systems which is a measure of topological order [70, 71, 79]. Could
the number h(D) be used as another index which characterizes the topological order?
January 23, 2014 95 Ali Nassar
Bibliography
[1] H. Georgi, Lie Algebras in Particle Physics. Westview Press, 1999.
[2] M. Gell-Mann and Y. Neemam, The Eightfold Way: A Review with a
Collection of Reprints (Frontiers in Physics Series: Lecture Note and Reprint
Series). W. A. Benjamin, Inc Publishers, 1964.
[3] C.-N. Yang and R. L. Mills, “Conservation of isotopic spin and isotopic gauge
invariance,” Phys. Rev. 96 (1954) 191–195.
[4] G. ’t Hooft, “Under the spell of the gauge principle,” Adv. Ser. Math. Phys. 19
(1994) 1–683.
[5] J. Wess and B. Zumino, “A Lagrangian model invariant under supergauge
transformations,” Phys. Lett. B49 (1974) 52.
[6] J. Wess and B. Zumino, “Supergauge transformations in four-dimensions,”
Nucl. Phys. B70 (1974) 39–50.
[7] A. Salam and J. A. Strathdee, “Supergauge transformations,” Nucl. Phys. B76
(1974) 477–482.
[8] A. M. Jaffe and E. Witten, Quantum Yang-Mills Theory. Clay Mathematics
Institute Millenium Prize Problem, 2000.
96
BIBLIOGRAPHY
[9] J. Polchinski, String Theory. Vol. 1: An Introduction to the Bosonic String.
Cambridge University Press, 1998.
[10] J. Polchinski, String Theory. Vol. 2: Superstring Theory and Beyond.
Cambridge University Press, 1998.
[11] R. Blumenhagen, B. Kors, D. Lust, and S. Stieberger, “Four-dimensional string
compactifications with D-branes, orientifolds and fluxes,” Phys. Rept. 445
(2007) 1–193, arXiv:hep-th/0610327 [hep-th].
[12] J. L. Cardy, “Conformal invariance and surface critical behavior,” Nucl. Phys.
B240 (1984) 514–532.
[13] J. L. Cardy, “Boundary conditions, fusion rules and the Verlinde formula,”
Nucl. Phys. B324 (1989) 581.
[14] J. Polchinski, “Dirichlet branes and Ramond-Ramond charges,” Phys. Rev.
Lett. 75 (1995) 4724–4727, arXiv:hep-th/9510017 [hep-th].
[15] E. Witten, “Search for a realistic Kaluza-Klein theory,” Nucl. Phys. B186
(1981) 412.
[16] D. R. Morrison and C. Vafa, “Compactifications of F theory on Calabi-Yau
threefolds. 1,” Nucl. Phys. B473 (1996) 74–92, arXiv:hep-th/9602114
[hep-th].
[17] D. R. Morrison and C. Vafa, “Compactifications of F theory on Calabi-Yau
threefolds. 2.,” Nucl. Phys. B476 (1996) 437–469, arXiv:hep-th/9603161
[hep-th].
January 23, 2014 97 Ali Nassar
BIBLIOGRAPHY
[18] D. J. Gross, J. A. Harvey, E. J. Martinec, and R. Rohm, “The heterotic
string,” Phys. Rev. Lett. 54 (1985) 502–505.
[19] D. J. Gross, J. A. Harvey, E. J. Martinec, and R. Rohm, “Heterotic string
theory. 1. The free heterotic string,” Nucl. Phys. B256 (1985) 253.
[20] D. J. Gross, J. A. Harvey, E. J. Martinec, and R. Rohm, “Heterotic string
theory. 2. The interacting heterotic string,” Nucl. Phys. B267 (1986) 75.
[21] E. Witten, “Nonabelian bosonization in two-dimensions,” Commun. Math.
Phys. 92 (1984) 455–472.
[22] R. Kallosh, A. D. Linde, D. A. Linde, and L. Susskind, “Gravity and global
symmetries,” Phys.Rev. D52 (1995) 912–935, arXiv:hep-th/9502069
[hep-th].
[23] K. S. Narain, “New heterotic string theories in uncompactified dimensions
< 10,” Phys. Lett. B169 (1986) 41.
[24] K. S. Narain, M. H. Sarmadi, and E. Witten, “A Note on toroidal
compactification of heterotic string theory,” Nucl. Phys. B279 (1987) 369.
[25] L. J. Dixon, “Some world sheet properties of superstring compactifications, on
orbifolds and otherwise,”.
[26] W. Lerche, C. Vafa, and N. P. Warner, “Chiral rings in N = 2 superconformal
theories,” Nucl. Phys. B324 (1989) 427.
[27] B. R. Greene and M. R. Plesser, “Duality in Calabi-Yau moduli space,” Nucl.
Phys. B338 (1990) 15–37.
January 23, 2014 98 Ali Nassar
BIBLIOGRAPHY
[28] D. Gepner, “Space-time supersymmetry in compactified string theory and
superconformal models,” Nucl. Phys. B296 (1988) 757.
[29] J. A. Harvey, G. W. Moore, and C. Vafa, “Quasicrystalline compactification,”
Nucl. Phys. B304 (1988) 269–290.
[30] K. Wendland, Moduli spaces of unitary conformal field theories. PhD thesis,
Bonn University. Ph.D. Thesis, 2000.
[31] G. W. Moore, “Arithmetic and attractors,” arXiv:hep-th/9807087 [hep-th].
[32] S. Gukov and C. Vafa, “Rational conformal field theories and complex
multiplication,” Commun. Math. Phys. 246 (2004) 181–210,
arXiv:hep-th/0203213 [hep-th].
[33] S. Hosono, B. H. Lian, K. Oguiso, and S.-T. Yau, “Classification of c = 2
rational conformal field theories via the Gauss product,” Commun. Math. Phys.
241 (2003) 245–286, arXiv:hep-th/0211230 [hep-th].
[34] T. Gannon, “U(1)m modular invariants, N = 2 minimal models, and the
quantum Hall effect,” Nucl. Phys. B491 (1997) 659–688,
arXiv:hep-th/9608063 [hep-th].
[35] S. Carlip and I. I. Kogan, “Three-dimensional topological field theories and
strings,” Mod. Phys. Lett. A6 (1991) 171–182.
[36] L. Cooper and I. I. Kogan, “Boundary conditions and heterotic construction in
topological membrane theory,” Phys. Lett. B383 (1996) 271–280,
arXiv:hep-th/9602062 [hep-th].
January 23, 2014 99 Ali Nassar
BIBLIOGRAPHY
[37] L. Cooper, I. I. Kogan, and K.-M. Lee, “String winding modes from charge
nonconservation in compact Chern-Simons theory,” Phys. Lett. B394 (1997)
67–74, arXiv:hep-th/9611107 [hep-th].
[38] A. A. Belavin, A. M. Polyakov, and A. B. Zamolodchikov, “Infinite conformal
symmetry in two-dimensional quantum field theory,” Nucl. Phys. B241 (1984)
333–380.
[39] P. Di Francesco, P. Mathieu, and D. Senechal, Conformal Field Theory.
Springer, 1997.
[40] P. H. Ginsparg, “Applied conformal field theory,” arXiv:hep-th/9108028
[hep-th].
[41] J. Polchinski, “Scale and conformal invariance in quantum field theory,” Nucl.
Phys. B303 (1988) 226.
[42] J. L. Cardy, Scaling and Renormalization in Statistical Physics. Cambridge
University Press, 1996.
[43] L. Onsager, “Crystal statistics. 1. A Two-dimensional model with an order
disorder transition,” Phys. Rev. 65 (1944) 117–149.
[44] J. Sethna, Statistical Mechanics: Entropy, Order Parameters and Complexity.
Oxford University Press, 2006.
[45] T. Gannon, “Monstrous moonshine and the classification of CFT,”
arXiv:math/9906167 [math].
January 23, 2014 100 Ali Nassar
BIBLIOGRAPHY
[46] J. L. Cardy, “Operator content of two-dimensional conformally invariant
theories,” Nucl. Phys. B270 (1986) 186–204.
[47] S. Lang, Elliptic Functions, vol. 112 of Graduate Texts in Mathematics.
Springer-Verlag, New York, second ed., 1987.
http://dx.doi.org/10.1007/978-1-4612-4752-4. With an appendix by J. Tate.
[48] G. W. Moore and N. Seiberg, “Taming the conformal zoo,” Phys. Lett. B220
(1989) 422.
[49] T. Banks, L. J. Dixon, D. Friedan, and E. J. Martinec, “Phenomenology and
conformal field theory or can string theory predict the weak mixing angle?,”
Nucl. Phys. B299 (1988) 613–626.
[50] P. Goddard and D. I. Olive, “Kac-Moody and Virasoro algebras in relation to
quantum physics,” Int.J.Mod.Phys. A1 (1986) 303.
[51] V. G. Kac, “Simple graded Lie algebras of
finite growth,” : Funct. Anal. Appl. 1 (1967) 328.
[52] R. V. Moody, “Lie algebras associated with generalized Cartan matrices,” Bull.
Amer. Math. Soc. 73 (1967) 217–221.
[53] D. Gepner and E. Witten, “String theory on group manifolds,” Nucl. Phys.
B278 (1986) 493.
[54] A. B. Zamolodchikov, “Infinite additional symmetries in two-dimensional
conformal quantum field theory,” Theor. Math. Phys. 65 (1985) 1205–1213.
January 23, 2014 101 Ali Nassar
BIBLIOGRAPHY
[55] P. Bouwknegt and K. Schoutens, “W symmetry in conformal field theory,”
Phys. Rept. 223 (1993) 183–276, arXiv:hep-th/9210010 [hep-th].
[56] V. G. Knizhnik and A. B. Zamolodchikov, “Current algebra and Wess-Zumino
model in two-dimensions,” Nucl. Phys. B247 (1984) 83–103.
[57] D. Friedan and S. H. Shenker, “The analytic geometry of two-dimensional
conformal field theory,” Nucl. Phys. B281 (1987) 509.
[58] K. Gawedzki and A. Kupiainen, “G/h conformal field theory from gauged
WZW model,” Phys. Lett. B215 (1988) 119–123.
[59] E. Witten, “On holomorphic factorization of WZW and coset models,”
Commun. Math. Phys. 144 (1992) 189–212.
[60] G. Anderson and G. W. Moore, “Rationality in conformal field theory,”
Commun. Math. Phys. 117 (1988) 441.
[61] A. Cappelli, C. Itzykson, and J. B. Zuber, “The ADE classification of minimal
and A(1)1 conformal invariant theories,” Commun. Math. Phys. 113 (1987) 1.
[62] A. Cappelli, C. Itzykson, and J. B. Zuber, “Modular invariant partition
functions in two-dimensions,” Nucl. Phys. B280 (1987) 445–465.
[63] T. Gannon, “The Cappelli-Itzykson-Zuber A-D-E classification,” Rev. Math.
Phys. 12 (2000) 739–748, arXiv:math/9902064 [math].
[64] A. Giveon, M. Porrati, and E. Rabinovici, “Target space duality in string
theory,” Phys.Rept. 244 (1994) 77–202, arXiv:hep-th/9401139 [hep-th].
January 23, 2014 102 Ali Nassar
BIBLIOGRAPHY
[65] I. Dolgachev, “Integral quadratic forms: applications to algebraic geometry
(after V. Nikulin),” in Bourbaki seminar, Vol. 1982/83, vol. 105 of Asterisque,
pp. 251–278. Soc. Math. France, Paris, 1983.
[66] D. Belov and G. W. Moore, “Classification of Abelian spin Chern-Simons
theories,” arXiv:hep-th/0505235 [hep-th].
[67] A. Cappelli and G. R. Zemba, “Modular invariant partition functions in the
quantum Hall effect,” Nucl. Phys. B490 (1997) 595–632,
arXiv:hep-th/9605127 [hep-th].
[68] E. Witten, “Quantum field theory and the Jones polynomial,” Commun. Math.
Phys. 121 (1989) 351.
[69] R. Dijkgraaf and E. Witten, “Topological gauge theories and group
cohomology,” Commun. Math. Phys. 129 (1990) 393.
[70] X. G. Wen, “Topological order in rigid states,” Int. J. Mod. Phys. B4 (1990)
239.
[71] X.-G. Wen, “Topological orders and Chern-Simons theory in strongly
correlated quantum liquid,” Int. J. Mod. Phys. B5 (1991) 1641–1648.
[72] R. B. Laughlin, “Anomalous quantum Hall effect: An incompressible quantum
fluid with fractionally charged excitations,” Phys. Rev. Lett. 50 (May, 1983)
1395–1398. http://link.aps.org/doi/10.1103/PhysRevLett.50.1395.
[73] D. A. Cox, Primes of the Form x2 + ny2. A Wiley-Interscience Publication.
John Wiley and Sons Inc., New York, 1989. Fermat, class field theory and
complex multiplication.
January 23, 2014 103 Ali Nassar
BIBLIOGRAPHY
[74] E. Witten, “Phases of N=2 theories in two-dimensions,” Nucl. Phys. B403
(1993) 159–222, arXiv:hep-th/9301042 [hep-th].
[75] K. Hori, A. Iqbal, and C. Vafa, “D-branes and mirror symmetry,”
arXiv:hep-th/0005247 [hep-th].
[76] P. Candelas, X. de la Ossa, and F. Rodriguez-Villegas, “Calabi-Yau manifolds
over finite fields. 1.,” arXiv:hep-th/0012233 [hep-th].
[77] A. Nassar and M. A. Walton, “Rational conformal field theory with matrix level
and strings on a torus,” Canadian Journal of Physics 9 (2013) , 1211.2728.
[78] C. Vafa, “Quantum symmetries of string vacua,” Mod. Phys. Lett. A4 (1989)
1615.
[79] X.-G. Wen, “Vacuum degeneracy of chiral spin states in compactified space,”
Phys. Rev. B40 (1989) 7387–7390.
[80] S. Elitzur, G. W. Moore, A. Schwimmer, and N. Seiberg, “Remarks on the
canonical quantization of the Chern-Simons-Witten theory,” Nucl. Phys. B326
(1989) 108.
[81] G. V. Dunne, R. Jackiw, and C. A. Trugenberger, “Chern-Simons theory in the
Schrodinger representation,” Annals Phys. 194 (1989) 197.
[82] R. Dijkgraaf and E. P. Verlinde, “Modular invariance and the fusion algebra,”
Nucl. Phys. Proc. Suppl. 5B (1988) 87–97.
[83] G. W. Moore and N. Seiberg, “Polynomial equations for rational conformal
field theories,” Phys. Lett. B212 (1988) 451.
January 23, 2014 104 Ali Nassar